Kan Chen

Yinjun Wu, Mayank Keoliya, Kan Chen et al.

Designing faithful yet accurate AI models is challenging, particularly in the field of individual treatment effect estimation (ITE). ITE prediction models deployed in critical settings such as healthcare should ideally be (i) accurate, and (ii) provide faithful explanations. However, current solutions are inadequate: state-of-the-art black-box models do not supply explanations, post-hoc explainers for black-box models lack faithfulness guarantees, and self-interpretable models greatly compromise accuracy. To address these issues, we propose DISCRET, a self-interpretable ITE framework that synthesizes faithful, rule-based explanations for each sample. A key insight behind DISCRET is that explanations can serve dually as database queries to identify similar subgroups of samples. We provide a novel RL algorithm to efficiently synthesize these explanations from a large search space. We evaluate DISCRET on diverse tasks involving tabular, image, and text data. DISCRET outperforms the best self-interpretable models and has accuracy comparable to the best black-box models while providing faithful explanations. DISCRET is available at https://github.com/wuyinjun-1993/DISCRET-ICML2024.

Alvin Wan, Xiaoliang Dai, Peizhao Zhang et al.

Differentiable Neural Architecture Search (DNAS) has demonstrated great success in designing state-of-the-art, efficient neural networks. However, DARTS-based DNAS's search space is small when compared to other search methods', since all candidate network layers must be explicitly instantiated in memory. To address this bottleneck, we propose a memory and computationally efficient DNAS variant: DMaskingNAS. This algorithm expands the search space by up to $10^{14}\times$ over conventional DNAS, supporting searches over spatial and channel dimensions that are otherwise prohibitively expensive: input resolution and number of filters. We propose a masking mechanism for feature map reuse, so that memory and computational costs stay nearly constant as the search space expands. Furthermore, we employ effective shape propagation to maximize per-FLOP or per-parameter accuracy. The searched FBNetV2s yield state-of-the-art performance when compared with all previous architectures. With up to 421$\times$ less search cost, DMaskingNAS finds models with 0.9% higher accuracy, 15% fewer FLOPs than MobileNetV3-Small; and with similar accuracy but 20% fewer FLOPs than Efficient-B0. Furthermore, our FBNetV2 outperforms MobileNetV3 by 2.6% in accuracy, with equivalent model size. FBNetV2 models are open-sourced at https://github.com/facebookresearch/mobile-vision.

Norman Mu, Jingwei Ji, Zhenpei Yang et al.

Many existing motion prediction approaches rely on symbolic perception outputs to generate agent trajectories, such as bounding boxes, road graph information and traffic lights. This symbolic representation is a high-level abstraction of the real world, which may render the motion prediction model vulnerable to perception errors (e.g., failures in detecting open-vocabulary obstacles) while missing salient information from the scene context (e.g., poor road conditions). An alternative paradigm is end-to-end learning from raw sensors. However, this approach suffers from the lack of interpretability and requires significantly more training resources. In this work, we propose tokenizing the visual world into a compact set of scene elements and then leveraging pre-trained image foundation models and LiDAR neural networks to encode all the scene elements in an open-vocabulary manner. The image foundation model enables our scene tokens to encode the general knowledge of the open world while the LiDAR neural network encodes geometry information. Our proposed representation can efficiently encode the multi-frame multi-modality observations with a few hundred tokens and is compatible with most transformer-based architectures. To evaluate our method, we have augmented Waymo Open Motion Dataset with camera embeddings. Experiments over Waymo Open Motion Dataset show that our approach leads to significant performance improvements over the state-of-the-art.

Arka Sadhu, Kan Chen, Ram Nevatia

Video Question Answering (VidQA) evaluation metrics have been limited to a single-word answer or selecting a phrase from a fixed set of phrases. These metrics limit the VidQA models' application scenario. In this work, we leverage semantic roles derived from video descriptions to mask out certain phrases, to introduce VidQAP which poses VidQA as a fill-in-the-phrase task. To enable evaluation of answer phrases, we compute the relative improvement of the predicted answer compared to an empty string. To reduce the influence of language bias in VidQA datasets, we retrieve a video having a different answer for the same question. To facilitate research, we construct ActivityNet-SRL-QA and Charades-SRL-QA and benchmark them by extending three vision-language models. We further perform extensive analysis and ablative studies to guide future work.

Yen-Cheng Liu, Chih-Yao Ma, Zijian He et al.

Semi-supervised learning, i.e., training networks with both labeled and unlabeled data, has made significant progress recently. However, existing works have primarily focused on image classification tasks and neglected object detection which requires more annotation effort. In this work, we revisit the Semi-Supervised Object Detection (SS-OD) and identify the pseudo-labeling bias issue in SS-OD. To address this, we introduce Unbiased Teacher, a simple yet effective approach that jointly trains a student and a gradually progressing teacher in a mutually-beneficial manner. Together with a class-balance loss to downweight overly confident pseudo-labels, Unbiased Teacher consistently improved state-of-the-art methods by significant margins on COCO-standard, COCO-additional, and VOC datasets. Specifically, Unbiased Teacher achieves 6.8 absolute mAP improvements against state-of-the-art method when using 1% of labeled data on MS-COCO, achieves around 10 mAP improvements against the supervised baseline when using only 0.5, 1, 2% of labeled data on MS-COCO.

Xiaoliang Dai, Alvin Wan, Peizhao Zhang et al.

Neural Architecture Search (NAS) yields state-of-the-art neural networks that outperform their best manually-designed counterparts. However, previous NAS methods search for architectures under one set of training hyper-parameters (i.e., a training recipe), overlooking superior architecture-recipe combinations. To address this, we present Neural Architecture-Recipe Search (NARS) to search both (a) architectures and (b) their corresponding training recipes, simultaneously. NARS utilizes an accuracy predictor that scores architecture and training recipes jointly, guiding both sample selection and ranking. Furthermore, to compensate for the enlarged search space, we leverage "free" architecture statistics (e.g., FLOP count) to pretrain the predictor, significantly improving its sample efficiency and prediction reliability. After training the predictor via constrained iterative optimization, we run fast evolutionary searches in just CPU minutes to generate architecture-recipe pairs for a variety of resource constraints, called FBNetV3. FBNetV3 makes up a family of state-of-the-art compact neural networks that outperform both automatically and manually-designed competitors. For example, FBNetV3 matches both EfficientNet and ResNeSt accuracy on ImageNet with up to 2.0x and 7.1x fewer FLOPs, respectively. Furthermore, FBNetV3 yields significant performance gains for downstream object detection tasks, improving mAP despite 18% fewer FLOPs and 34% fewer parameters than EfficientNet-based equivalents.

Chuanzi He, Haidong Zhu, Jiyang Gao et al.

The task of referring relationships is to localize subject and object entities in an image satisfying a relationship query, which is given in the form of \texttt{<subject, predicate, object>}. This requires simultaneous localization of the subject and object entities in a specified relationship. We introduce a simple yet effective proposal-based method for referring relationships. Different from the existing methods such as SSAS, our method can generate a high-resolution result while reducing its complexity and ambiguity. Our method is composed of two modules: a category-based proposal generation module to select the proposals related to the entities and a predicate analysis module to score the compatibility of pairs of selected proposals. We show state-of-the-art performance on the referring relationship task on two public datasets: Visual Relationship Detection and Visual Genome.

Arka Sadhu, Kan Chen, Ram Nevatia

We explore the task of Video Object Grounding (VOG), which grounds objects in videos referred to in natural language descriptions. Previous methods apply image grounding based algorithms to address VOG, fail to explore the object relation information and suffer from limited generalization. Here, we investigate the role of object relations in VOG and propose a novel framework VOGNet to encode multi-modal object relations via self-attention with relative position encoding. To evaluate VOGNet, we propose novel contrasting sampling methods to generate more challenging grounding input samples, and construct a new dataset called ActivityNet-SRL (ASRL) based on existing caption and grounding datasets. Experiments on ASRL validate the need of encoding object relations in VOG, and our VOGNet outperforms competitive baselines by a significant margin.

Arka Sadhu, Kan Chen, Ram Nevatia

A phrase grounding system localizes a particular object in an image referred to by a natural language query. In previous work, the phrases were restricted to have nouns that were encountered in training, we extend the task to Zero-Shot Grounding(ZSG) which can include novel, "unseen" nouns. Current phrase grounding systems use an explicit object detection network in a 2-stage framework where one stage generates sparse proposals and the other stage evaluates them. In the ZSG setting, generating appropriate proposals itself becomes an obstacle as the proposal generator is trained on the entities common in the detection and grounding datasets. We propose a new single-stage model called ZSGNet which combines the detector network and the grounding system and predicts classification scores and regression parameters. Evaluation of ZSG system brings additional subtleties due to the influence of the relationship between the query and learned categories; we define four distinct conditions that incorporate different levels of difficulty. We also introduce new datasets, sub-sampled from Flickr30k Entities and Visual Genome, that enable evaluations for the four conditions. Our experiments show that ZSGNet achieves state-of-the-art performance on Flickr30k and ReferIt under the usual "seen" settings and performs significantly better than baseline in the zero-shot setting.

I. Zeki Yalniz, Hervé Jégou, Kan Chen et al.

This paper presents a study of semi-supervised learning with large convolutional networks. We propose a pipeline, based on a teacher/student paradigm, that leverages a large collection of unlabelled images (up to 1 billion). Our main goal is to improve the performance for a given target architecture, like ResNet-50 or ResNext. We provide an extensive analysis of the success factors of our approach, which leads us to formulate some recommendations to produce high-accuracy models for image classification with semi-supervised learning. As a result, our approach brings important gains to standard architectures for image, video and fine-grained classification. For instance, by leveraging one billion unlabelled images, our learned vanilla ResNet-50 achieves 81.2% top-1 accuracy on the ImageNet benchmark.

Runzhou Ge, Jiyang Gao, Kan Chen et al.

We address the problem of language-based temporal localization in untrimmed videos. Compared to temporal localization with fixed categories, this problem is more challenging as the language-based queries not only have no pre-defined activity list but also may contain complex descriptions. Previous methods address the problem by considering features from video sliding windows and language queries and learning a subspace to encode their correlation, which ignore rich semantic cues about activities in videos and queries. We propose to mine activity concepts from both video and language modalities by applying the actionness score enhanced Activity Concepts based Localizer (ACL). Specifically, the novel ACL encodes the semantic concepts from verb-obj pairs in language queries and leverages activity classifiers' prediction scores to encode visual concepts. Besides, ACL also has the capability to regress sliding windows as localization results. Experiments show that ACL significantly outperforms state-of-the-arts under the widely used metric, with more than 5% increase on both Charades-STA and TACoS datasets.

Jiyang Gao, Kan Chen, Ram Nevatia

Temporal action proposal generation is an important task, akin to object proposals, temporal action proposals are intended to capture "clips" or temporal intervals in videos that are likely to contain an action. Previous methods can be divided to two groups: sliding window ranking and actionness score grouping. Sliding windows uniformly cover all segments in videos, but the temporal boundaries are imprecise; grouping based method may have more precise boundaries but it may omit some proposals when the quality of actionness score is low. Based on the complementary characteristics of these two methods, we propose a novel Complementary Temporal Action Proposal (CTAP) generator. Specifically, we apply a Proposal-level Actionness Trustworthiness Estimator (PATE) on the sliding windows proposals to generate the probabilities indicating whether the actions can be correctly detected by actionness scores, the windows with high scores are collected. The collected sliding windows and actionness proposals are then processed by a temporal convolutional neural network for proposal ranking and boundary adjustment. CTAP outperforms state-of-the-art methods on average recall (AR) by a large margin on THUMOS-14 and ActivityNet 1.3 datasets. We further apply CTAP as a proposal generation method in an existing action detector, and show consistent significant improvements.

Jiyang Gao, Runzhou Ge, Kan Chen et al.

Video Question Answering (QA) is an important task in understanding video temporal structure. We observe that there are three unique attributes of video QA compared with image QA: (1) it deals with long sequences of images containing richer information not only in quantity but also in variety; (2) motion and appearance information are usually correlated with each other and able to provide useful attention cues to the other; (3) different questions require different number of frames to infer the answer. Based these observations, we propose a motion-appearance comemory network for video QA. Our networks are built on concepts from Dynamic Memory Network (DMN) and introduces new mechanisms for video QA. Specifically, there are three salient aspects: (1) a co-memory attention mechanism that utilizes cues from both motion and appearance to generate attention; (2) a temporal conv-deconv network to generate multi-level contextual facts; (3) a dynamic fact ensemble method to construct temporal representation dynamically for different questions. We evaluate our method on TGIF-QA dataset, and the results outperform state-of-the-art significantly on all four tasks of TGIF-QA.

Kan Chen, Jiyang Gao, Ram Nevatia

Given a natural language query, a phrase grounding system aims to localize mentioned objects in an image. In weakly supervised scenario, mapping between image regions (i.e., proposals) and language is not available in the training set. Previous methods address this deficiency by training a grounding system via learning to reconstruct language information contained in input queries from predicted proposals. However, the optimization is solely guided by the reconstruction loss from the language modality, and ignores rich visual information contained in proposals and useful cues from external knowledge. In this paper, we explore the consistency contained in both visual and language modalities, and leverage complementary external knowledge to facilitate weakly supervised grounding. We propose a novel Knowledge Aided Consistency Network (KAC Net) which is optimized by reconstructing input query and proposal's information. To leverage complementary knowledge contained in the visual features, we introduce a Knowledge Based Pooling (KBP) gate to focus on query-related proposals. Experiments show that KAC Net provides a significant improvement on two popular datasets.

Kan Chen, Rama Kovvuri, Ram Nevatia

Given a textual description of an image, phrase grounding localizes objects in the image referred by query phrases in the description. State-of-the-art methods address the problem by ranking a set of proposals based on the relevance to each query, which are limited by the performance of independent proposal generation systems and ignore useful cues from context in the description. In this paper, we adopt a spatial regression method to break the performance limit, and introduce reinforcement learning techniques to further leverage semantic context information. We propose a novel Query-guided Regression network with Context policy (QRC Net) which jointly learns a Proposal Generation Network (PGN), a Query-guided Regression Network (QRN) and a Context Policy Network (CPN). Experiments show QRC Net provides a significant improvement in accuracy on two popular datasets: Flickr30K Entities and Referit Game, with 14.25% and 17.14% increase over the state-of-the-arts respectively.

Kan Chen, Trung Bui, Fang Chen et al.

Given a user's query, traditional image search systems rank images according to its relevance to a single modality (e.g., image content or surrounding text). Nowadays, an increasing number of images on the Internet are available with associated meta data in rich modalities (e.g., titles, keywords, tags, etc.), which can be exploited for better similarity measure with queries. In this paper, we leverage visual and textual modalities for image search by learning their correlation with input query. According to the intent of query, attention mechanism can be introduced to adaptively balance the importance of different modalities. We propose a novel Attention guided Multi-modal Correlation (AMC) learning method which consists of a jointly learned hierarchy of intra and inter-attention networks. Conditioned on query's intent, intra-attention networks (i.e., visual intra-attention network and language intra-attention network) attend on informative parts within each modality; a multi-modal inter-attention network promotes the importance of the most query-relevant modalities. In experiments, we evaluate AMC models on the search logs from two real world image search engines and show a significant boost on the ranking of user-clicked images in search results. Additionally, we extend AMC models to caption ranking task on COCO dataset and achieve competitive results compared with recent state-of-the-arts.

Jiyang Gao, Zhenheng Yang, Chen Sun et al.

Temporal Action Proposal (TAP) generation is an important problem, as fast and accurate extraction of semantically important (e.g. human actions) segments from untrimmed videos is an important step for large-scale video analysis. We propose a novel Temporal Unit Regression Network (TURN) model. There are two salient aspects of TURN: (1) TURN jointly predicts action proposals and refines the temporal boundaries by temporal coordinate regression; (2) Fast computation is enabled by unit feature reuse: a long untrimmed video is decomposed into video units, which are reused as basic building blocks of temporal proposals. TURN outperforms the state-of-the-art methods under average recall (AR) by a large margin on THUMOS-14 and ActivityNet datasets, and runs at over 880 frames per second (FPS) on a TITAN X GPU. We further apply TURN as a proposal generation stage for existing temporal action localization pipelines, it outperforms state-of-the-art performance on THUMOS-14 and ActivityNet.

Kan Chen, Jiang Wang, Liang-Chieh Chen et al.

We propose a novel attention based deep learning architecture for visual question answering task (VQA). Given an image and an image related natural language question, VQA generates the natural language answer for the question. Generating the correct answers requires the model's attention to focus on the regions corresponding to the question, because different questions inquire about the attributes of different image regions. We introduce an attention based configurable convolutional neural network (ABC-CNN) to learn such question-guided attention. ABC-CNN determines an attention map for an image-question pair by convolving the image feature map with configurable convolutional kernels derived from the question's semantics. We evaluate the ABC-CNN architecture on three benchmark VQA datasets: Toronto COCO-QA, DAQUAR, and VQA dataset. ABC-CNN model achieves significant improvements over state-of-the-art methods on these datasets. The question-guided attention generated by ABC-CNN is also shown to reflect the regions that are highly relevant to the questions.