Jonathon Shlens

Wentao Wu, Aleksei Timofeev, Chen Chen et al. · microsoft-research

We present MOFI, Manifold OF Images, a new vision foundation model designed to learn image representations from noisy entity annotated images. MOFI differs from previous work in two key aspects: (i) pre-training data, and (ii) training recipe. Regarding data, we introduce a new approach to automatically assign entity labels to images from noisy image-text pairs. Our approach involves employing a named entity recognition model to extract entities from the alt-text, and then using a CLIP model to select the correct entities as labels of the paired image. It's a simple, cost-effective method that can scale to handle billions of web-mined image-text pairs. Through this method, we have created Image-to-Entities (I2E), a new dataset with 1 billion images and 2 million distinct entities, covering rich visual concepts in the wild. Building upon the I2E dataset, we study different training recipes like supervised pre-training, contrastive pre-training, and multi-task learning. For contrastive pre-training, we treat entity names as free-form text, and further enrich them with entity descriptions. Experiments show that supervised pre-training with large-scale fine-grained entity labels is highly effective for image retrieval tasks, and multi-task training further improves the performance. The final MOFI model achieves 86.66% mAP on the challenging GPR1200 dataset, surpassing the previous state-of-the-art performance of 72.19% from OpenAI's CLIP model. Further experiments on zero-shot and linear probe image classification also show that MOFI outperforms a CLIP model trained on the original image-text data, demonstrating the effectiveness of the I2E dataset in learning strong image representations. We release our code and model weights at https://github.com/apple/ml-mofi.

Kanchana Ranasinghe, Brandon McKinzie, Sachin Ravi et al.

Recent advances in zero-shot image recognition suggest that vision-language models learn generic visual representations with a high degree of semantic information that may be arbitrarily probed with natural language phrases. Understanding an image, however, is not just about understanding what content resides within an image, but importantly, where that content resides. In this work we examine how well vision-language models are able to understand where objects reside within an image and group together visually related parts of the imagery. We demonstrate how contemporary vision and language representation learning models based on contrastive losses and large web-based data capture limited object localization information. We propose a minimal set of modifications that results in models that uniquely learn both semantic and spatial information. We measure this performance in terms of zero-shot image recognition, unsupervised bottom-up and top-down semantic segmentations, as well as robustness analyses. We find that the resulting model achieves state-of-the-art results in terms of unsupervised segmentation, and demonstrate that the learned representations are uniquely robust to spurious correlations in datasets designed to probe the causal behavior of vision models.

Chen Chen, Bowen Zhang, Liangliang Cao et al.

Image and text retrieval is one of the foundational tasks in the vision and language domain with multiple real-world applications. State-of-the-art approaches, e.g. CLIP, ALIGN, represent images and texts as dense embeddings and calculate the similarity in the dense embedding space as the matching score. On the other hand, sparse semantic features like bag-of-words models are more interpretable, but believed to suffer from inferior accuracy than dense representations. In this work, we show that it is possible to build a sparse semantic representation that is as powerful as, or even better than, dense presentations. We extend the CLIP model and build a sparse text and image representation (STAIR), where the image and text are mapped to a sparse token space. Each token in the space is a (sub-)word in the vocabulary, which is not only interpretable but also easy to integrate with existing information retrieval systems. STAIR model significantly outperforms a CLIP model with +$4.9\%$ and +$4.3\%$ absolute Recall@1 improvement on COCO-5k text$\rightarrow$image and image$\rightarrow$text retrieval respectively. It also achieved better performance on both of ImageNet zero-shot and linear probing compared to CLIP.

Zhaoqi Leng, Shuyang Cheng, Benjamin Caine et al.

Data augmentation is an important technique to improve data efficiency and save labeling cost for 3D detection in point clouds. Yet, existing augmentation policies have so far been designed to only utilize labeled data, which limits the data diversity. In this paper, we recognize that pseudo labeling and data augmentation are complementary, thus propose to leverage unlabeled data for data augmentation to enrich the training data. In particular, we design three novel pseudo-label based data augmentation policies (PseudoAugments) to fuse both labeled and pseudo-labeled scenes, including frames (PseudoFrame), objecta (PseudoBBox), and background (PseudoBackground). PseudoAugments outperforms pseudo labeling by mitigating pseudo labeling errors and generating diverse fused training scenes. We demonstrate PseudoAugments generalize across point-based and voxel-based architectures, different model capacity and both KITTI and Waymo Open Dataset. To alleviate the cost of hyperparameter tuning and iterative pseudo labeling, we develop a population-based data augmentation framework for 3D detection, named AutoPseudoAugment. Unlike previous works that perform pseudo-labeling offline, our framework performs PseudoAugments and hyperparameter tuning in one shot to reduce computational cost. Experimental results on the large-scale Waymo Open Dataset show our method outperforms state-of-the-art auto data augmentation method (PPBA) and self-training method (pseudo labeling). In particular, AutoPseudoAugment is about 3X and 2X data efficient on vehicle and pedestrian tasks compared to prior arts. Notably, AutoPseudoAugment nearly matches the full dataset training results, with just 10% of the labeled run segments on the vehicle detection task.

Brandon McKinzie, Joseph Cheng, Vaishaal Shankar et al.

Multimodal learning is defined as learning over multiple heterogeneous input modalities such as video, audio, and text. In this work, we are concerned with understanding how models behave as the type of modalities differ between training and deployment, a situation that naturally arises in many applications of multimodal learning to hardware platforms. We present a multimodal robustness framework to provide a systematic analysis of common multimodal representation learning methods. Further, we identify robustness short-comings of these approaches and propose two intervention techniques leading to $1.5\times$-$4\times$ robustness improvements on three datasets, AudioSet, Kinetics-400 and ImageNet-Captions. Finally, we demonstrate that these interventions better utilize additional modalities, if present, to achieve competitive results of $44.2$ mAP on AudioSet 20K.

Robert Geirhos, Priyank Jaini, Austin Stone et al.

Training a neural network is a monolithic endeavor, akin to carving knowledge into stone: once the process is completed, editing the knowledge in a network is hard, since all information is distributed across the network's weights. We here explore a simple, compelling alternative by marrying the representational power of deep neural networks with the flexibility of a database. Decomposing the task of image classification into image similarity (from a pre-trained embedding) and search (via fast nearest neighbor retrieval from a knowledge database), we build on well-established components to construct a simple and flexible visual memory that has the following key capabilities: (1.) The ability to flexibly add data across scales: from individual samples all the way to entire classes and billion-scale data; (2.) The ability to remove data through unlearning and memory pruning; (3.) An interpretable decision-mechanism on which we can intervene to control its behavior. Taken together, these capabilities comprehensively demonstrate the benefits of an explicit visual memory. We hope that it might contribute to a conversation on how knowledge should be represented in deep vision models -- beyond carving it in "stone" weights.

Khaled Saab, Tao Tu, Wei-Hung Weng et al.

Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-Gemini, a family of highly capable multimodal models that are specialized in medicine with the ability to seamlessly use web search, and that can be efficiently tailored to novel modalities using custom encoders. We evaluate Med-Gemini on 14 medical benchmarks, establishing new state-of-the-art (SoTA) performance on 10 of them, and surpass the GPT-4 model family on every benchmark where a direct comparison is viable, often by a wide margin. On the popular MedQA (USMLE) benchmark, our best-performing Med-Gemini model achieves SoTA performance of 91.1% accuracy, using a novel uncertainty-guided search strategy. On 7 multimodal benchmarks including NEJM Image Challenges and MMMU (health & medicine), Med-Gemini improves over GPT-4V by an average relative margin of 44.5%. We demonstrate the effectiveness of Med-Gemini's long-context capabilities through SoTA performance on a needle-in-a-haystack retrieval task from long de-identified health records and medical video question answering, surpassing prior bespoke methods using only in-context learning. Finally, Med-Gemini's performance suggests real-world utility by surpassing human experts on tasks such as medical text summarization, alongside demonstrations of promising potential for multimodal medical dialogue, medical research and education. Taken together, our results offer compelling evidence for Med-Gemini's potential, although further rigorous evaluation will be crucial before real-world deployment in this safety-critical domain.

Ekin D. Cubuk, Barret Zoph, Jonathon Shlens et al.

Recent work has shown that data augmentation has the potential to significantly improve the generalization of deep learning models. Recently, automated augmentation strategies have led to state-of-the-art results in image classification and object detection. While these strategies were optimized for improving validation accuracy, they also led to state-of-the-art results in semi-supervised learning and improved robustness to common corruptions of images. An obstacle to a large-scale adoption of these methods is a separate search phase which increases the training complexity and may substantially increase the computational cost. Additionally, due to the separate search phase, these approaches are unable to adjust the regularization strength based on model or dataset size. Automated augmentation policies are often found by training small models on small datasets and subsequently applied to train larger models. In this work, we remove both of these obstacles. RandAugment has a significantly reduced search space which allows it to be trained on the target task with no need for a separate proxy task. Furthermore, due to the parameterization, the regularization strength may be tailored to different model and dataset sizes. RandAugment can be used uniformly across different tasks and datasets and works out of the box, matching or surpassing all previous automated augmentation approaches on CIFAR-10/100, SVHN, and ImageNet. On the ImageNet dataset we achieve 85.0% accuracy, a 0.6% increase over the previous state-of-the-art and 1.0% increase over baseline augmentation. On object detection, RandAugment leads to 1.0-1.3% improvement over baseline augmentation, and is within 0.3% mAP of AutoAugment on COCO. Finally, due to its interpretable hyperparameter, RandAugment may be used to investigate the role of data augmentation with varying model and dataset size. Code is available online.

Barret Zoph, Ekin D. Cubuk, Golnaz Ghiasi et al.

Data augmentation is a critical component of training deep learning models. Although data augmentation has been shown to significantly improve image classification, its potential has not been thoroughly investigated for object detection. Given the additional cost for annotating images for object detection, data augmentation may be of even greater importance for this computer vision task. In this work, we study the impact of data augmentation on object detection. We first demonstrate that data augmentation operations borrowed from image classification may be helpful for training detection models, but the improvement is limited. Thus, we investigate how learned, specialized data augmentation policies improve generalization performance for detection models. Importantly, these augmentation policies only affect training and leave a trained model unchanged during evaluation. Experiments on the COCO dataset indicate that an optimized data augmentation policy improves detection accuracy by more than +2.3 mAP, and allow a single inference model to achieve a state-of-the-art accuracy of 50.7 mAP. Importantly, the best policy found on COCO may be transferred unchanged to other detection datasets and models to improve predictive accuracy. For example, the best augmentation policy identified with COCO improves a strong baseline on PASCAL-VOC by +2.7 mAP. Our results also reveal that a learned augmentation policy is superior to state-of-the-art architecture regularization methods for object detection, even when considering strong baselines. Code for training with the learned policy is available online at https://github.com/tensorflow/tpu/tree/master/models/official/detection

Martín Abadi, Ashish Agarwal, Paul Barham et al.

TensorFlow is an interface for expressing machine learning algorithms, and an implementation for executing such algorithms. A computation expressed using TensorFlow can be executed with little or no change on a wide variety of heterogeneous systems, ranging from mobile devices such as phones and tablets up to large-scale distributed systems of hundreds of machines and thousands of computational devices such as GPU cards. The system is flexible and can be used to express a wide variety of algorithms, including training and inference algorithms for deep neural network models, and it has been used for conducting research and for deploying machine learning systems into production across more than a dozen areas of computer science and other fields, including speech recognition, computer vision, robotics, information retrieval, natural language processing, geographic information extraction, and computational drug discovery. This paper describes the TensorFlow interface and an implementation of that interface that we have built at Google. The TensorFlow API and a reference implementation were released as an open-source package under the Apache 2.0 license in November, 2015 and are available at www.tensorflow.org.

Andrew Sellergren, Sahar Kazemzadeh, Tiam Jaroensri et al.

Artificial intelligence (AI) has significant potential in healthcare applications, but its training and deployment faces challenges due to healthcare's diverse data, complex tasks, and the need to preserve privacy. Foundation models that perform well on medical tasks and require less task-specific tuning data are critical to accelerate the development of healthcare AI applications. We introduce MedGemma, a collection of medical vision-language foundation models based on Gemma 3 4B and 27B. MedGemma demonstrates advanced medical understanding and reasoning on images and text, significantly exceeding the performance of similar-sized generative models and approaching the performance of task-specific models, while maintaining the general capabilities of the Gemma 3 base models. For out-of-distribution tasks, MedGemma achieves 2.6-10% improvement on medical multimodal question answering, 15.5-18.1% improvement on chest X-ray finding classification, and 10.8% improvement on agentic evaluations compared to the base models. Fine-tuning MedGemma further improves performance in subdomains, reducing errors in electronic health record information retrieval by 50% and reaching comparable performance to existing specialized state-of-the-art methods for pneumothorax classification and histopathology patch classification. We additionally introduce MedSigLIP, a medically-tuned vision encoder derived from SigLIP. MedSigLIP powers the visual understanding capabilities of MedGemma and as an encoder achieves comparable or better performance than specialized medical image encoders. Taken together, the MedGemma collection provides a strong foundation of medical image and text capabilities, with potential to significantly accelerate medical research and development of downstream applications. The MedGemma collection, including tutorials and model weights, can be found at https://goo.gle/medgemma.

Austin Stone, Hagen Soltau, Robert Geirhos et al.

Visual imagery does not consist of solitary objects, but instead reflects the composition of a multitude of fluid concepts. While there have been great advances in visual representation learning, such advances have focused on building better representations for a small number of discrete objects bereft of an understanding of how these objects are interacting. One can observe this limitation in representations learned through captions or contrastive learning -- where the learned model treats an image essentially as a bag of words. Several works have attempted to address this limitation through the development of bespoke learned architectures to directly address the shortcomings in compositional learning. In this work, we focus on simple, and scalable approaches. In particular, we demonstrate that by substantially improving weakly labeled data, i.e. captions, we can vastly improve the performance of standard contrastive learning approaches. Previous CLIP models achieved near chance rate on challenging tasks probing compositional learning. However, our simple approach boosts performance of CLIP substantially and surpasses all bespoke architectures. Furthermore, we showcase our results on a relatively new captioning benchmark derived from DOCCI. We demonstrate through a series of ablations that a standard CLIP model trained with enhanced data may demonstrate impressive performance on image retrieval tasks.

Archit Karandikar, Nicholas Cain, Dustin Tran et al.

Optimal decision making requires that classifiers produce uncertainty estimates consistent with their empirical accuracy. However, deep neural networks are often under- or over-confident in their predictions. Consequently, methods have been developed to improve the calibration of their predictive uncertainty both during training and post-hoc. In this work, we propose differentiable losses to improve calibration based on a soft (continuous) version of the binning operation underlying popular calibration-error estimators. When incorporated into training, these soft calibration losses achieve state-of-the-art single-model ECE across multiple datasets with less than 1% decrease in accuracy. For instance, we observe an 82% reduction in ECE (70% relative to the post-hoc rescaled ECE) in exchange for a 0.7% relative decrease in accuracy relative to the cross entropy baseline on CIFAR-100. When incorporated post-training, the soft-binning-based calibration error objective improves upon temperature scaling, a popular recalibration method. Overall, experiments across losses and datasets demonstrate that using calibration-sensitive procedures yield better uncertainty estimates under dataset shift than the standard practice of using a cross entropy loss and post-hoc recalibration methods.

Jiquan Ngiam, Benjamin Caine, Vijay Vasudevan et al.

Predicting the motion of multiple agents is necessary for planning in dynamic environments. This task is challenging for autonomous driving since agents (e.g. vehicles and pedestrians) and their associated behaviors may be diverse and influence one another. Most prior work have focused on predicting independent futures for each agent based on all past motion, and planning against these independent predictions. However, planning against independent predictions can make it challenging to represent the future interaction possibilities between different agents, leading to sub-optimal planning. In this work, we formulate a model for predicting the behavior of all agents jointly, producing consistent futures that account for interactions between agents. Inspired by recent language modeling approaches, we use a masking strategy as the query to our model, enabling one to invoke a single model to predict agent behavior in many ways, such as potentially conditioned on the goal or full future trajectory of the autonomous vehicle or the behavior of other agents in the environment. Our model architecture employs attention to combine features across road elements, agent interactions, and time steps. We evaluate our approach on autonomous driving datasets for both marginal and joint motion prediction, and achieve state of the art performance across two popular datasets. Through combining a scene-centric approach, agent permutation equivariant model, and a sequence masking strategy, we show that our model can unify a variety of motion prediction tasks from joint motion predictions to conditioned prediction.

Scott Ettinger, Shuyang Cheng, Benjamin Caine et al.

As autonomous driving systems mature, motion forecasting has received increasing attention as a critical requirement for planning. Of particular importance are interactive situations such as merges, unprotected turns, etc., where predicting individual object motion is not sufficient. Joint predictions of multiple objects are required for effective route planning. There has been a critical need for high-quality motion data that is rich in both interactions and annotation to develop motion planning models. In this work, we introduce the most diverse interactive motion dataset to our knowledge, and provide specific labels for interacting objects suitable for developing joint prediction models. With over 100,000 scenes, each 20 seconds long at 10 Hz, our new dataset contains more than 570 hours of unique data over 1750 km of roadways. It was collected by mining for interesting interactions between vehicles, pedestrians, and cyclists across six cities within the United States. We use a high-accuracy 3D auto-labeling system to generate high quality 3D bounding boxes for each road agent, and provide corresponding high definition 3D maps for each scene. Furthermore, we introduce a new set of metrics that provides a comprehensive evaluation of both single agent and joint agent interaction motion forecasting models. Finally, we provide strong baseline models for individual-agent prediction and joint-prediction. We hope that this new large-scale interactive motion dataset will provide new opportunities for advancing motion forecasting models.

Ashish Vaswani, Prajit Ramachandran, Aravind Srinivas et al.

Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we aim to develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutional models.

Irwan Bello, William Fedus, Xianzhi Du et al.

Novel computer vision architectures monopolize the spotlight, but the impact of the model architecture is often conflated with simultaneous changes to training methodology and scaling strategies. Our work revisits the canonical ResNet (He et al., 2015) and studies these three aspects in an effort to disentangle them. Perhaps surprisingly, we find that training and scaling strategies may matter more than architectural changes, and further, that the resulting ResNets match recent state-of-the-art models. We show that the best performing scaling strategy depends on the training regime and offer two new scaling strategies: (1) scale model depth in regimes where overfitting can occur (width scaling is preferable otherwise); (2) increase image resolution more slowly than previously recommended (Tan & Le, 2019). Using improved training and scaling strategies, we design a family of ResNet architectures, ResNet-RS, which are 1.7x - 2.7x faster than EfficientNets on TPUs, while achieving similar accuracies on ImageNet. In a large-scale semi-supervised learning setup, ResNet-RS achieves 86.2% top-1 ImageNet accuracy, while being 4.7x faster than EfficientNet NoisyStudent. The training techniques improve transfer performance on a suite of downstream tasks (rivaling state-of-the-art self-supervised algorithms) and extend to video classification on Kinetics-400. We recommend practitioners use these simple revised ResNets as baselines for future research.

Benjamin Caine, Rebecca Roelofs, Vijay Vasudevan et al.

To safely deploy autonomous vehicles, onboard perception systems must work reliably at high accuracy across a diverse set of environments and geographies. One of the most common techniques to improve the efficacy of such systems in new domains involves collecting large labeled datasets, but such datasets can be extremely costly to obtain, especially if each new deployment geography requires additional data with expensive 3D bounding box annotations. We demonstrate that pseudo-labeling for 3D object detection is an effective way to exploit less expensive and more widely available unlabeled data, and can lead to performance gains across various architectures, data augmentation strategies, and sizes of the labeled dataset. Overall, we show that better teacher models lead to better student models, and that we can distill expensive teachers into efficient, simple students. Specifically, we demonstrate that pseudo-label-trained student models can outperform supervised models trained on 3-10 times the amount of labeled examples. Using PointPillars [24], a two-year-old architecture, as our student model, we are able to achieve state of the art accuracy simply by leveraging large quantities of pseudo-labeled data. Lastly, we show that these student models generalize better than supervised models to a new domain in which we only have unlabeled data, making pseudo-label training an effective form of unsupervised domain adaptation.

Philipp Jund, Chris Sweeney, Nichola Abdo et al.

Autonomous vehicles operate in highly dynamic environments necessitating an accurate assessment of which aspects of a scene are moving and where they are moving to. A popular approach to 3D motion estimation, termed scene flow, is to employ 3D point cloud data from consecutive LiDAR scans, although such approaches have been limited by the small size of real-world, annotated LiDAR data. In this work, we introduce a new large-scale dataset for scene flow estimation derived from corresponding tracked 3D objects, which is $\sim$1,000$\times$ larger than previous real-world datasets in terms of the number of annotated frames. We demonstrate how previous works were bounded based on the amount of real LiDAR data available, suggesting that larger datasets are required to achieve state-of-the-art predictive performance. Furthermore, we show how previous heuristics for operating on point clouds such as down-sampling heavily degrade performance, motivating a new class of models that are tractable on the full point cloud. To address this issue, we introduce the FastFlow3D architecture which provides real time inference on the full point cloud. Additionally, we design human-interpretable metrics that better capture real world aspects by accounting for ego-motion and providing breakdowns per object type. We hope that this dataset may provide new opportunities for developing real world scene flow systems.

Aravind Srinivas, Tsung-Yi Lin, Niki Parmar et al.

We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 1.64x faster in compute time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision

Rebecca Roelofs, Nicholas Cain, Jonathon Shlens et al.

For an AI system to be reliable, the confidence it expresses in its decisions must match its accuracy. To assess the degree of match, examples are typically binned by confidence and the per-bin mean confidence and accuracy are compared. Most research in calibration focuses on techniques to reduce this empirical measure of calibration error, ECE_bin. We instead focus on assessing statistical bias in this empirical measure, and we identify better estimators. We propose a framework through which we can compute the bias of a particular estimator for an evaluation data set of a given size. The framework involves synthesizing model outputs that have the same statistics as common neural architectures on popular data sets. We find that binning-based estimators with bins of equal mass (number of instances) have lower bias than estimators with bins of equal width. Our results indicate two reliable calibration-error estimators: the debiased estimator (Brocker, 2012; Ferro and Fricker, 2012) and a method we propose, ECE_sweep, which uses equal-mass bins and chooses the number of bins to be as large as possible while preserving monotonicity in the calibration function. With these estimators, we observe improvements in the effectiveness of recalibration methods and in the detection of model miscalibration.

Liang-Chieh Chen, Raphael Gontijo Lopes, Bowen Cheng et al.

Supervised learning in large discriminative models is a mainstay for modern computer vision. Such an approach necessitates investing in large-scale human-annotated datasets for achieving state-of-the-art results. In turn, the efficacy of supervised learning may be limited by the size of the human annotated dataset. This limitation is particularly notable for image segmentation tasks, where the expense of human annotation is especially large, yet large amounts of unlabeled data may exist. In this work, we ask if we may leverage semi-supervised learning in unlabeled video sequences and extra images to improve the performance on urban scene segmentation, simultaneously tackling semantic, instance, and panoptic segmentation. The goal of this work is to avoid the construction of sophisticated, learned architectures specific to label propagation (e.g., patch matching and optical flow). Instead, we simply predict pseudo-labels for the unlabeled data and train subsequent models with both human-annotated and pseudo-labeled data. The procedure is iterated for several times. As a result, our Naive-Student model, trained with such simple yet effective iterative semi-supervised learning, attains state-of-the-art results at all three Cityscapes benchmarks, reaching the performance of 67.8% PQ, 42.6% AP, and 85.2% mIOU on the test set. We view this work as a notable step towards building a simple procedure to harness unlabeled video sequences and extra images to surpass state-of-the-art performance on core computer vision tasks.

Wei Han, Zhengdong Zhang, Benjamin Caine et al.

Autonomous vehicles operate in a dynamic environment, where the speed with which a vehicle can perceive and react impacts the safety and efficacy of the system. LiDAR provides a prominent sensory modality that informs many existing perceptual systems including object detection, segmentation, motion estimation, and action recognition. The latency for perceptual systems based on point cloud data can be dominated by the amount of time for a complete rotational scan (e.g. 100 ms). This built-in data capture latency is artificial, and based on treating the point cloud as a camera image in order to leverage camera-inspired architectures. However, unlike camera sensors, most LiDAR point cloud data is natively a streaming data source in which laser reflections are sequentially recorded based on the precession of the laser beam. In this work, we explore how to build an object detector that removes this artificial latency constraint, and instead operates on native streaming data in order to significantly reduce latency. This approach has the added benefit of reducing the peak computational burden on inference hardware by spreading the computation over the acquisition time for a scan. We demonstrate a family of streaming detection systems based on sequential modeling through a series of modifications to the traditional detection meta-architecture. We highlight how this model may achieve competitive if not superior predictive performance with state-of-the-art, traditional non-streaming detection systems while achieving significant latency gains (e.g. 1/15'th - 1/3'rd of peak latency). Our results show that operating on LiDAR data in its native streaming formulation offers several advantages for self driving object detection -- advantages that we hope will be useful for any LiDAR perception system where minimizing latency is critical for safe and efficient operation.

Shuyang Cheng, Zhaoqi Leng, Ekin Dogus Cubuk et al.

Data augmentation has been widely adopted for object detection in 3D point clouds. However, all previous related efforts have focused on manually designing specific data augmentation methods for individual architectures. In this work, we present the first attempt to automate the design of data augmentation policies for 3D object detection. We introduce the Progressive Population Based Augmentation (PPBA) algorithm, which learns to optimize augmentation strategies by narrowing down the search space and adopting the best parameters discovered in previous iterations. On the KITTI 3D detection test set, PPBA improves the StarNet detector by substantial margins on the moderate difficulty category of cars, pedestrians, and cyclists, outperforming all current state-of-the-art single-stage detection models. Additional experiments on the Waymo Open Dataset indicate that PPBA continues to effectively improve the StarNet and PointPillars detectors on a 20x larger dataset compared to KITTI. The magnitude of the improvements may be comparable to advances in 3D perception architectures and the gains come without an incurred cost at inference time. In subsequent experiments, we find that PPBA may be up to 10x more data efficient than baseline 3D detection models without augmentation, highlighting that 3D detection models may achieve competitive accuracy with far fewer labeled examples.

Gamaleldin F. Elsayed, Prajit Ramachandran, Jonathon Shlens et al.

Convolutional neural networks are among the most successful architectures in deep learning with this success at least partially attributable to the efficacy of spatial invariance as an inductive bias. Locally connected layers, which differ from convolutional layers only in their lack of spatial invariance, usually perform poorly in practice. However, these observations still leave open the possibility that some degree of relaxation of spatial invariance may yield a better inductive bias than either convolution or local connectivity. To test this hypothesis, we design a method to relax the spatial invariance of a network layer in a controlled manner; we create a \textit{low-rank} locally connected layer, where the filter bank applied at each position is constructed as a linear combination of basis set of filter banks with spatially varying combining weights. By varying the number of basis filter banks, we can control the degree of relaxation of spatial invariance. In experiments with small convolutional networks, we find that relaxing spatial invariance improves classification accuracy over both convolution and locally connected layers across MNIST, CIFAR-10, and CelebA datasets, thus suggesting that spatial invariance may be an overly restrictive prior.

Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla et al.

The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the overall viability of the technology. In an effort to help align the research community's contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open.

Jiquan Ngiam, Benjamin Caine, Wei Han et al.

Detecting objects from LiDAR point clouds is an important component of self-driving car technology as LiDAR provides high resolution spatial information. Previous work on point-cloud 3D object detection has re-purposed convolutional approaches from traditional camera imagery. In this work, we present an object detection system called StarNet designed specifically to take advantage of the sparse and 3D nature of point cloud data. StarNet is entirely point-based, uses no global information, has data dependent anchors, and uses sampling instead of learned region proposals. We demonstrate how this design leads to competitive or superior performance on the large Waymo Open Dataset and the KITTI detection dataset, as compared to convolutional baselines. In particular, we show how our detector can outperform a competitive baseline on Pedestrian detection on the Waymo Open Dataset by more than 7 absolute mAP while being more computationally efficient. We show how our redesign---namely using only local information and using sampling instead of learned proposals---leads to a significantly more flexible and adaptable system: we demonstrate how we can vary the computational cost of a single trained StarNet without retraining, and how we can target proposals towards areas of interest with priors and heuristics. Finally, we show how our design allows for incorporating temporal context by using detections from previous frames to target computation of the detector, which leads to further improvements in performance without additional computational cost.

Dong Yin, Raphael Gontijo Lopes, Jonathon Shlens et al.

Achieving robustness to distributional shift is a longstanding and challenging goal of computer vision. Data augmentation is a commonly used approach for improving robustness, however robustness gains are typically not uniform across corruption types. Indeed increasing performance in the presence of random noise is often met with reduced performance on other corruptions such as contrast change. Understanding when and why these sorts of trade-offs occur is a crucial step towards mitigating them. Towards this end, we investigate recently observed trade-offs caused by Gaussian data augmentation and adversarial training. We find that both methods improve robustness to corruptions that are concentrated in the high frequency domain while reducing robustness to corruptions that are concentrated in the low frequency domain. This suggests that one way to mitigate these trade-offs via data augmentation is to use a more diverse set of augmentations. Towards this end we observe that AutoAugment, a recently proposed data augmentation policy optimized for clean accuracy, achieves state-of-the-art robustness on the CIFAR-10-C benchmark.

Prajit Ramachandran, Niki Parmar, Ashish Vaswani et al.

Convolutions are a fundamental building block of modern computer vision systems. Recent approaches have argued for going beyond convolutions in order to capture long-range dependencies. These efforts focus on augmenting convolutional models with content-based interactions, such as self-attention and non-local means, to achieve gains on a number of vision tasks. The natural question that arises is whether attention can be a stand-alone primitive for vision models instead of serving as just an augmentation on top of convolutions. In developing and testing a pure self-attention vision model, we verify that self-attention can indeed be an effective stand-alone layer. A simple procedure of replacing all instances of spatial convolutions with a form of self-attention applied to ResNet model produces a fully self-attentional model that outperforms the baseline on ImageNet classification with 12% fewer FLOPS and 29% fewer parameters. On COCO object detection, a pure self-attention model matches the mAP of a baseline RetinaNet while having 39% fewer FLOPS and 34% fewer parameters. Detailed ablation studies demonstrate that self-attention is especially impactful when used in later layers. These results establish that stand-alone self-attention is an important addition to the vision practitioner's toolbox.

Aakanksha Chowdhery, Pete Warden, Jonathon Shlens et al.

The emergence of Internet of Things (IoT) applications requires intelligence on the edge. Microcontrollers provide a low-cost compute platform to deploy intelligent IoT applications using machine learning at scale, but have extremely limited on-chip memory and compute capability. To deploy computer vision on such devices, we need tiny vision models that fit within a few hundred kilobytes of memory footprint in terms of peak usage and model size on device storage. To facilitate the development of microcontroller friendly models, we present a new dataset, Visual Wake Words, that represents a common microcontroller vision use-case of identifying whether a person is present in the image or not, and provides a realistic benchmark for tiny vision models. Within a limited memory footprint of 250 KB, several state-of-the-art mobile models achieve accuracy of 85-90% on the Visual Wake Words dataset. We anticipate the proposed dataset will advance the research on tiny vision models that can push the pareto-optimal boundary in terms of accuracy versus memory usage for microcontroller applications.

Luke Metz, Niru Maheswaranathan, Jonathon Shlens et al.

State-of-the art vision models can achieve superhuman performance on image classification tasks when testing and training data come from the same distribution. However, when models are tested on corrupted images (e.g. due to scale changes, translations, or shifts in brightness or contrast), performance degrades significantly. Here, we explore the possibility of meta-training a learned optimizer that can train image classification models such that they are robust to common image corruptions. Specifically, we are interested training models that are more robust to noise distributions not present in the training data. We find that a learned optimizer meta-trained to produce models which are robust to Gaussian noise trains models that are more robust to Gaussian noise at other scales compared to traditional optimizers like Adam. The effect of meta-training is more complicated when targeting a more general set of noise distributions, but led to improved performance on half of held-out corruption tasks. Our results suggest that meta-learning provides a novel approach for studying and improving the robustness of deep learning models.

Keren Gu, Brandon Yang, Jiquan Ngiam et al.

Human visual systems are robust to a wide range of image transformations that are challenging for artificial networks. We present the first study of image model robustness to the minute transformations found across video frames, which we term "natural robustness". Compared to previous studies on adversarial examples and synthetic distortions, natural robustness captures a more diverse set of common image transformations that occur in the natural environment. Our study across a dozen model architectures shows that more accurate models are more robust to natural transformations, and that robustness to synthetic color distortions is a good proxy for natural robustness. In examining brittleness in videos, we find that majority of the brittleness found in videos lies outside the typical definition of adversarial examples (99.9\%). Finally, we investigate training techniques to reduce brittleness and find that no single technique systematically improves natural robustness across twelve tested architectures.

Irwan Bello, Barret Zoph, Ashish Vaswani et al.

Convolutional networks have been the paradigm of choice in many computer vision applications. The convolution operation however has a significant weakness in that it only operates on a local neighborhood, thus missing global information. Self-attention, on the other hand, has emerged as a recent advance to capture long range interactions, but has mostly been applied to sequence modeling and generative modeling tasks. In this paper, we consider the use of self-attention for discriminative visual tasks as an alternative to convolutions. We introduce a novel two-dimensional relative self-attention mechanism that proves competitive in replacing convolutions as a stand-alone computational primitive for image classification. We find in control experiments that the best results are obtained when combining both convolutions and self-attention. We therefore propose to augment convolutional operators with this self-attention mechanism by concatenating convolutional feature maps with a set of feature maps produced via self-attention. Extensive experiments show that Attention Augmentation leads to consistent improvements in image classification on ImageNet and object detection on COCO across many different models and scales, including ResNets and a state-of-the art mobile constrained network, while keeping the number of parameters similar. In particular, our method achieves a $1.3\%$ top-1 accuracy improvement on ImageNet classification over a ResNet50 baseline and outperforms other attention mechanisms for images such as Squeeze-and-Excitation. It also achieves an improvement of 1.4 mAP in COCO Object Detection on top of a RetinaNet baseline.

Raphael Gontijo Lopes, David Ha, Douglas Eck et al.

Dramatic advances in generative models have resulted in near photographic quality for artificially rendered faces, animals and other objects in the natural world. In spite of such advances, a higher level understanding of vision and imagery does not arise from exhaustively modeling an object, but instead identifying higher-level attributes that best summarize the aspects of an object. In this work we attempt to model the drawing process of fonts by building sequential generative models of vector graphics. This model has the benefit of providing a scale-invariant representation for imagery whose latent representation may be systematically manipulated and exploited to perform style propagation. We demonstrate these results on a large dataset of fonts and highlight how such a model captures the statistical dependencies and richness of this dataset. We envision that our model can find use as a tool for graphic designers to facilitate font design.

Jasmine Collins, Johannes Balle, Jonathon Shlens

A significant advance in accelerating neural network training has been the development of normalization methods, permitting the training of deep models both faster and with better accuracy. These advances come with practical challenges: for instance, batch normalization ties the prediction of individual examples with other examples within a batch, resulting in a network that is heavily dependent on batch size. Layer normalization and group normalization are data-dependent and thus must be continually used, even at test-time. To address the issues that arise from using explicit normalization techniques, we propose to replace existing normalization methods with a simple, secondary objective loss that we term a standardization loss. This formulation is flexible and robust across different batch sizes and surprisingly, this secondary objective accelerates learning on the primary training objective. Because it is a training loss, it is simply removed at test-time, and no further effort is needed to maintain normalized activations. We find that a standardization loss accelerates training on both small- and large-scale image classification experiments, works with a variety of architectures, and is largely robust to training across different batch sizes.

Liang-Chieh Chen, Maxwell D. Collins, Yukun Zhu et al.

The design of neural network architectures is an important component for achieving state-of-the-art performance with machine learning systems across a broad array of tasks. Much work has endeavored to design and build architectures automatically through clever construction of a search space paired with simple learning algorithms. Recent progress has demonstrated that such meta-learning methods may exceed scalable human-invented architectures on image classification tasks. An open question is the degree to which such methods may generalize to new domains. In this work we explore the construction of meta-learning techniques for dense image prediction focused on the tasks of scene parsing, person-part segmentation, and semantic image segmentation. Constructing viable search spaces in this domain is challenging because of the multi-scale representation of visual information and the necessity to operate on high resolution imagery. Based on a survey of techniques in dense image prediction, we construct a recursive search space and demonstrate that even with efficient random search, we can identify architectures that outperform human-invented architectures and achieve state-of-the-art performance on three dense prediction tasks including 82.7\% on Cityscapes (street scene parsing), 71.3\% on PASCAL-Person-Part (person-part segmentation), and 87.9\% on PASCAL VOC 2012 (semantic image segmentation). Additionally, the resulting architecture is more computationally efficient, requiring half the parameters and half the computational cost as previous state of the art systems.

Simon Kornblith, Jonathon Shlens, Quoc V. Le

Transfer learning is a cornerstone of computer vision, yet little work has been done to evaluate the relationship between architecture and transfer. An implicit hypothesis in modern computer vision research is that models that perform better on ImageNet necessarily perform better on other vision tasks. However, this hypothesis has never been systematically tested. Here, we compare the performance of 16 classification networks on 12 image classification datasets. We find that, when networks are used as fixed feature extractors or fine-tuned, there is a strong correlation between ImageNet accuracy and transfer accuracy ($r = 0.99$ and $0.96$, respectively). In the former setting, we find that this relationship is very sensitive to the way in which networks are trained on ImageNet; many common forms of regularization slightly improve ImageNet accuracy but yield penultimate layer features that are much worse for transfer learning. Additionally, we find that, on two small fine-grained image classification datasets, pretraining on ImageNet provides minimal benefits, indicating the learned features from ImageNet do not transfer well to fine-grained tasks. Together, our results show that ImageNet architectures generalize well across datasets, but ImageNet features are less general than previously suggested.

Guangyu Robert Yang, Igor Ganichev, Xiao-Jing Wang et al.

A vexing problem in artificial intelligence is reasoning about events that occur in complex, changing visual stimuli such as in video analysis or game play. Inspired by a rich tradition of visual reasoning and memory in cognitive psychology and neuroscience, we developed an artificial, configurable visual question and answer dataset (COG) to parallel experiments in humans and animals. COG is much simpler than the general problem of video analysis, yet it addresses many of the problems relating to visual and logical reasoning and memory -- problems that remain challenging for modern deep learning architectures. We additionally propose a deep learning architecture that performs competitively on other diagnostic VQA datasets (i.e. CLEVR) as well as easy settings of the COG dataset. However, several settings of COG result in datasets that are progressively more challenging to learn. After training, the network can zero-shot generalize to many new tasks. Preliminary analyses of the network architectures trained on COG demonstrate that the network accomplishes the task in a manner interpretable to humans.

Chenxi Liu, Barret Zoph, Maxim Neumann et al.

We propose a new method for learning the structure of convolutional neural networks (CNNs) that is more efficient than recent state-of-the-art methods based on reinforcement learning and evolutionary algorithms. Our approach uses a sequential model-based optimization (SMBO) strategy, in which we search for structures in order of increasing complexity, while simultaneously learning a surrogate model to guide the search through structure space. Direct comparison under the same search space shows that our method is up to 5 times more efficient than the RL method of Zoph et al. (2018) in terms of number of models evaluated, and 8 times faster in terms of total compute. The structures we discover in this way achieve state of the art classification accuracies on CIFAR-10 and ImageNet.

Lane McIntosh, Niru Maheswaranathan, David Sussillo et al.

State-of-the-art systems for semantic image segmentation use feed-forward pipelines with fixed computational costs. Building an image segmentation system that works across a range of computational budgets is challenging and time-intensive as new architectures must be designed and trained for every computational setting. To address this problem we develop a recurrent neural network that successively improves prediction quality with each iteration. Importantly, the RNN may be deployed across a range of computational budgets by merely running the model for a variable number of iterations. We find that this architecture is uniquely suited for efficiently segmenting videos. By exploiting the segmentation of past frames, the RNN can perform video segmentation at similar quality but reduced computational cost compared to state-of-the-art image segmentation methods. When applied to static images in the PASCAL VOC 2012 and Cityscapes segmentation datasets, the RNN traces out a speed-accuracy curve that saturates near the performance of state-of-the-art segmentation methods.

Barret Zoph, Vijay Vasudevan, Jonathon Shlens et al.

Developing neural network image classification models often requires significant architecture engineering. In this paper, we study a method to learn the model architectures directly on the dataset of interest. As this approach is expensive when the dataset is large, we propose to search for an architectural building block on a small dataset and then transfer the block to a larger dataset. The key contribution of this work is the design of a new search space (the "NASNet search space") which enables transferability. In our experiments, we search for the best convolutional layer (or "cell") on the CIFAR-10 dataset and then apply this cell to the ImageNet dataset by stacking together more copies of this cell, each with their own parameters to design a convolutional architecture, named "NASNet architecture". We also introduce a new regularization technique called ScheduledDropPath that significantly improves generalization in the NASNet models. On CIFAR-10 itself, NASNet achieves 2.4% error rate, which is state-of-the-art. On ImageNet, NASNet achieves, among the published works, state-of-the-art accuracy of 82.7% top-1 and 96.2% top-5 on ImageNet. Our model is 1.2% better in top-1 accuracy than the best human-invented architectures while having 9 billion fewer FLOPS - a reduction of 28% in computational demand from the previous state-of-the-art model. When evaluated at different levels of computational cost, accuracies of NASNets exceed those of the state-of-the-art human-designed models. For instance, a small version of NASNet also achieves 74% top-1 accuracy, which is 3.1% better than equivalently-sized, state-of-the-art models for mobile platforms. Finally, the learned features by NASNet used with the Faster-RCNN framework surpass state-of-the-art by 4.0% achieving 43.1% mAP on the COCO dataset.

Sergio Guadarrama, Ryan Dahl, David Bieber et al.

We propose a novel approach to automatically produce multiple colorized versions of a grayscale image. Our method results from the observation that the task of automated colorization is relatively easy given a low-resolution version of the color image. We first train a conditional PixelCNN to generate a low resolution color for a given grayscale image. Then, given the generated low-resolution color image and the original grayscale image as inputs, we train a second CNN to generate a high-resolution colorization of an image. We demonstrate that our approach produces more diverse and plausible colorizations than existing methods, as judged by human raters in a "Visual Turing Test".

Golnaz Ghiasi, Honglak Lee, Manjunath Kudlur et al.

In this paper, we present a method which combines the flexibility of the neural algorithm of artistic style with the speed of fast style transfer networks to allow real-time stylization using any content/style image pair. We build upon recent work leveraging conditional instance normalization for multi-style transfer networks by learning to predict the conditional instance normalization parameters directly from a style image. The model is successfully trained on a corpus of roughly 80,000 paintings and is able to generalize to paintings previously unobserved. We demonstrate that the learned embedding space is smooth and contains a rich structure and organizes semantic information associated with paintings in an entirely unsupervised manner.

Esteban Real, Jonathon Shlens, Stefano Mazzocchi et al.

We introduce a new large-scale data set of video URLs with densely-sampled object bounding box annotations called YouTube-BoundingBoxes (YT-BB). The data set consists of approximately 380,000 video segments about 19s long, automatically selected to feature objects in natural settings without editing or post-processing, with a recording quality often akin to that of a hand-held cell phone camera. The objects represent a subset of the MS COCO label set. All video segments were human-annotated with high-precision classification labels and bounding boxes at 1 frame per second. The use of a cascade of increasingly precise human annotations ensures a label accuracy above 95% for every class and tight bounding boxes. Finally, we train and evaluate well-known deep network architectures and report baseline figures for per-frame classification and localization to provide a point of comparison for future work. We also demonstrate how the temporal contiguity of video can potentially be used to improve such inferences. Please see the PDF file to find the URL to download the data. We hope the availability of such large curated corpus will spur new advances in video object detection and tracking.

Ryan Dahl, Mohammad Norouzi, Jonathon Shlens

We present a pixel recursive super resolution model that synthesizes realistic details into images while enhancing their resolution. A low resolution image may correspond to multiple plausible high resolution images, thus modeling the super resolution process with a pixel independent conditional model often results in averaging different details--hence blurry edges. By contrast, our model is able to represent a multimodal conditional distribution by properly modeling the statistical dependencies among the high resolution image pixels, conditioned on a low resolution input. We employ a PixelCNN architecture to define a strong prior over natural images and jointly optimize this prior with a deep conditioning convolutional network. Human evaluations indicate that samples from our proposed model look more photo realistic than a strong L2 regression baseline.

Augustus Odena, Christopher Olah, Jonathon Shlens

Synthesizing high resolution photorealistic images has been a long-standing challenge in machine learning. In this paper we introduce new methods for the improved training of generative adversarial networks (GANs) for image synthesis. We construct a variant of GANs employing label conditioning that results in 128x128 resolution image samples exhibiting global coherence. We expand on previous work for image quality assessment to provide two new analyses for assessing the discriminability and diversity of samples from class-conditional image synthesis models. These analyses demonstrate that high resolution samples provide class information not present in low resolution samples. Across 1000 ImageNet classes, 128x128 samples are more than twice as discriminable as artificially resized 32x32 samples. In addition, 84.7% of the classes have samples exhibiting diversity comparable to real ImageNet data.

Vincent Dumoulin, Jonathon Shlens, Manjunath Kudlur

The diversity of painting styles represents a rich visual vocabulary for the construction of an image. The degree to which one may learn and parsimoniously capture this visual vocabulary measures our understanding of the higher level features of paintings, if not images in general. In this work we investigate the construction of a single, scalable deep network that can parsimoniously capture the artistic style of a diversity of paintings. We demonstrate that such a network generalizes across a diversity of artistic styles by reducing a painting to a point in an embedding space. Importantly, this model permits a user to explore new painting styles by arbitrarily combining the styles learned from individual paintings. We hope that this work provides a useful step towards building rich models of paintings and offers a window on to the structure of the learned representation of artistic style.

Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe et al.

Convolutional networks are at the core of most state-of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we explore ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21.2% top-1 and 5.6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3.5% top-5 error on the validation set (3.6% error on the test set) and 17.3% top-1 error on the validation set.

Alireza Makhzani, Jonathon Shlens, Navdeep Jaitly et al.

In this paper, we propose the "adversarial autoencoder" (AAE), which is a probabilistic autoencoder that uses the recently proposed generative adversarial networks (GAN) to perform variational inference by matching the aggregated posterior of the hidden code vector of the autoencoder with an arbitrary prior distribution. Matching the aggregated posterior to the prior ensures that generating from any part of prior space results in meaningful samples. As a result, the decoder of the adversarial autoencoder learns a deep generative model that maps the imposed prior to the data distribution. We show how the adversarial autoencoder can be used in applications such as semi-supervised classification, disentangling style and content of images, unsupervised clustering, dimensionality reduction and data visualization. We performed experiments on MNIST, Street View House Numbers and Toronto Face datasets and show that adversarial autoencoders achieve competitive results in generative modeling and semi-supervised classification tasks.

Tianqi Chen, Ian Goodfellow, Jonathon Shlens

We introduce techniques for rapidly transferring the information stored in one neural net into another neural net. The main purpose is to accelerate the training of a significantly larger neural net. During real-world workflows, one often trains very many different neural networks during the experimentation and design process. This is a wasteful process in which each new model is trained from scratch. Our Net2Net technique accelerates the experimentation process by instantaneously transferring the knowledge from a previous network to each new deeper or wider network. Our techniques are based on the concept of function-preserving transformations between neural network specifications. This differs from previous approaches to pre-training that altered the function represented by a neural net when adding layers to it. Using our knowledge transfer mechanism to add depth to Inception modules, we demonstrate a new state of the art accuracy rating on the ImageNet dataset.