Byungseok Roh

Taehoon Kim, Pyunghwan Ahn, Sangyun Kim et al. · nvidia, utoronto

In this report, we introduce NICE (New frontiers for zero-shot Image Captioning Evaluation) project and share the results and outcomes of 2023 challenge. This project is designed to challenge the computer vision community to develop robust image captioning models that advance the state-of-the-art both in terms of accuracy and fairness. Through the challenge, the image captioning models were tested using a new evaluation dataset that includes a large variety of visual concepts from many domains. There was no specific training data provided for the challenge, and therefore the challenge entries were required to adapt to new types of image descriptions that had not been seen during training. This report includes information on the newly proposed NICE dataset, evaluation methods, challenge results, and technical details of top-ranking entries. We expect that the outcomes of the challenge will contribute to the improvement of AI models on various vision-language tasks.

Junbum Cha, Jonghwan Mun, Byungseok Roh

We tackle open-world semantic segmentation, which aims at learning to segment arbitrary visual concepts in images, by using only image-text pairs without dense annotations. Existing open-world segmentation methods have shown impressive advances by employing contrastive learning (CL) to learn diverse visual concepts and transferring the learned image-level understanding to the segmentation task. However, these CL-based methods suffer from a train-test discrepancy, since it only considers image-text alignment during training, whereas segmentation requires region-text alignment during testing. In this paper, we proposed a novel Text-grounded Contrastive Learning (TCL) framework that enables a model to directly learn region-text alignment. Our method generates a segmentation mask for a given text, extracts text-grounded image embedding from the masked region, and aligns it with text embedding via TCL. By learning region-text alignment directly, our framework encourages a model to directly improve the quality of generated segmentation masks. In addition, for a rigorous and fair comparison, we present a unified evaluation protocol with widely used 8 semantic segmentation datasets. TCL achieves state-of-the-art zero-shot segmentation performances with large margins in all datasets. Code is available at https://github.com/kakaobrain/tcl.

Dohwan Ko, Joonmyung Choi, Hyeong Kyu Choi et al.

Foundation models have shown outstanding performance and generalization capabilities across domains. Since most studies on foundation models mainly focus on the pretraining phase, a naive strategy to minimize a single task-specific loss is adopted for fine-tuning. However, such fine-tuning methods do not fully leverage other losses that are potentially beneficial for the target task. Therefore, we propose MEta Loss TRansformer (MELTR), a plug-in module that automatically and non-linearly combines various loss functions to aid learning the target task via auxiliary learning. We formulate the auxiliary learning as a bi-level optimization problem and present an efficient optimization algorithm based on Approximate Implicit Differentiation (AID). For evaluation, we apply our framework to various video foundation models (UniVL, Violet and All-in-one), and show significant performance gain on all four downstream tasks: text-to-video retrieval, video question answering, video captioning, and multi-modal sentiment analysis. Our qualitative analyses demonstrate that MELTR adequately `transforms' individual loss functions and `melts' them into an effective unified loss. Code is available at https://github.com/mlvlab/MELTR.

Wooyoung Kang, Jonghwan Mun, Sungjun Lee et al.

Image captioning is one of the straightforward tasks that can take advantage of large-scale web-crawled data which provides rich knowledge about the visual world for a captioning model. However, since web-crawled data contains image-text pairs that are aligned at different levels, the inherent noises (e.g., misaligned pairs) make it difficult to learn a precise captioning model. While the filtering strategy can effectively remove noisy data, it leads to a decrease in learnable knowledge and sometimes brings about a new problem of data deficiency. To take the best of both worlds, we propose a Noise-aware Captioning (NoC) framework, which learns rich knowledge from the whole web-crawled data while being less affected by the noises. This is achieved by the proposed alignment-level-controllable captioner, which is learned using alignment levels of the image-text pairs as a control signal during training. The alignment-level-conditioned training allows the model to generate high-quality captions by simply setting the control signal to the desired alignment level at inference time. An in-depth analysis shows the effectiveness of our framework in handling noise. With two tasks of zero-shot captioning and text-to-image retrieval using generated captions (i.e., self-retrieval), we also demonstrate our model can produce high-quality captions in terms of descriptiveness and distinctiveness. The code is available at \url{https://github.com/kakaobrain/noc}.

Kihyun You, Jawook Gu, Jiyeon Ham et al.

A large-scale image-text pair dataset has greatly contributed to the development of vision-language pre-training (VLP) models, which enable zero-shot or few-shot classification without costly annotation. However, in the medical domain, the scarcity of data remains a significant challenge for developing a powerful VLP model. In this paper, we tackle the lack of image-text data in chest X-ray by expanding image-label pair as image-text pair via general prompt and utilizing multiple images and multiple sections in a radiologic report. We also design two contrastive losses, named ICL and TCL, for learning study-level characteristics of medical images and reports, respectively. Our model outperforms the state-of-the-art models trained under the same conditions. Also, enlarged dataset improve the discriminative power of our pre-trained model for classification, while sacrificing marginal retrieval performance. Code is available at https://github.com/kakaobrain/cxr-clip.

Dohwan Ko, Ji Soo Lee, Wooyoung Kang et al.

Large Language Models (LLMs) have shown remarkable performances on a wide range of natural language understanding and generation tasks. We observe that the LLMs provide effective priors in exploiting $\textit{linguistic shortcuts}$ for temporal and causal reasoning in Video Question Answering (VideoQA). However, such priors often cause suboptimal results on VideoQA by leading the model to over-rely on questions, $\textit{i.e.}$, $\textit{linguistic bias}$, while ignoring visual content. This is also known as `ungrounded guesses' or `hallucinations'. To address this problem while leveraging LLMs' prior on VideoQA, we propose a novel framework, Flipped-VQA, encouraging the model to predict all the combinations of $\langle$V, Q, A$\rangle$ triplet by flipping the source pair and the target label to understand their complex relationships, $\textit{i.e.}$, predict A, Q, and V given a VQ, VA, and QA pairs, respectively. In this paper, we develop LLaMA-VQA by applying Flipped-VQA to LLaMA, and it outperforms both LLMs-based and non-LLMs-based models on five challenging VideoQA benchmarks. Furthermore, our Flipped-VQA is a general framework that is applicable to various LLMs (OPT and GPT-J) and consistently improves their performances. We empirically demonstrate that Flipped-VQA not only enhances the exploitation of linguistic shortcuts but also mitigates the linguistic bias, which causes incorrect answers over-relying on the question. Code is available at https://github.com/mlvlab/Flipped-VQA.

Junbum Cha, Wooyoung Kang, Jonghwan Mun et al.

In Multimodal Large Language Models (MLLMs), a visual projector plays a crucial role in bridging pre-trained vision encoders with LLMs, enabling profound visual understanding while harnessing the LLMs' robust capabilities. Despite the importance of the visual projector, it has been relatively less explored. In this study, we first identify two essential projector properties: (i) flexibility in managing the number of visual tokens, crucial for MLLMs' overall efficiency, and (ii) preservation of local context from visual features, vital for spatial understanding. Based on these findings, we propose a novel projector design that is both flexible and locality-enhanced, effectively satisfying the two desirable properties. Additionally, we present comprehensive strategies to effectively utilize multiple and multifaceted instruction datasets. Through extensive experiments, we examine the impact of individual design choices. Finally, our proposed MLLM, Honeybee, remarkably outperforms previous state-of-the-art methods across various benchmarks, including MME, MMBench, SEED-Bench, and LLaVA-Bench, achieving significantly higher efficiency. Code and models are available at https://github.com/kakaobrain/honeybee.

Han-Cheol Cho, Won Young Jhoo, Wooyoung Kang et al.

Recent open-vocabulary detection methods aim to detect novel objects by distilling knowledge from vision-language models (VLMs) trained on a vast amount of image-text pairs. To improve the effectiveness of these methods, researchers have utilized datasets with a large vocabulary that contains a large number of object classes, under the assumption that such data will enable models to extract comprehensive knowledge on the relationships between various objects and better generalize to unseen object classes. In this study, we argue that more fine-grained labels are necessary to extract richer knowledge about novel objects, including object attributes and relationships, in addition to their names. To address this challenge, we propose a simple and effective method named Pseudo Caption Labeling (PCL), which utilizes an image captioning model to generate captions that describe object instances from diverse perspectives. The resulting pseudo caption labels offer dense samples for knowledge distillation. On the LVIS benchmark, our best model trained on the de-duplicated VisualGenome dataset achieves an AP of 34.5 and an APr of 30.6, comparable to the state-of-the-art performance. PCL's simplicity and flexibility are other notable features, as it is a straightforward pre-processing technique that can be used with any image captioning model without imposing any restrictions on model architecture or training process.

Jawook Gu, Kihyun You, Han-Cheol Cho et al.

Free-text radiology reports present a rich data source for various medical tasks, but effectively labeling these texts remains challenging. Traditional rule-based labeling methods fall short of capturing the nuances of diverse free-text patterns. Moreover, models using expert-annotated data are limited by data scarcity and pre-defined classes, impacting their performance, flexibility and scalability. To address these issues, our study offers three main contributions: 1) We demonstrate the potential of GPT as an adept labeler using carefully designed prompts. 2) Utilizing only the data labeled by GPT, we trained a BERT-based labeler, CheX-GPT, which operates faster and more efficiently than its GPT counterpart. 3) To benchmark labeler performance, we introduced a publicly available expert-annotated test set, MIMIC-500, comprising 500 cases from the MIMIC validation set. Our findings demonstrate that CheX-GPT not only excels in labeling accuracy over existing models, but also showcases superior efficiency, flexibility, and scalability, supported by our introduction of the MIMIC-500 dataset for robust benchmarking. Code and models are available at https://github.com/Soombit-ai/CheXGPT.

Byungseok Roh, JaeWoong Shin, Wuhyun Shin et al.

DETR is the first end-to-end object detector using a transformer encoder-decoder architecture and demonstrates competitive performance but low computational efficiency on high resolution feature maps. The subsequent work, Deformable DETR, enhances the efficiency of DETR by replacing dense attention with deformable attention, which achieves 10x faster convergence and improved performance. Deformable DETR uses the multiscale feature to ameliorate performance, however, the number of encoder tokens increases by 20x compared to DETR, and the computation cost of the encoder attention remains a bottleneck. In our preliminary experiment, we observe that the detection performance hardly deteriorates even if only a part of the encoder token is updated. Inspired by this observation, we propose Sparse DETR that selectively updates only the tokens expected to be referenced by the decoder, thus help the model effectively detect objects. In addition, we show that applying an auxiliary detection loss on the selected tokens in the encoder improves the performance while minimizing computational overhead. We validate that Sparse DETR achieves better performance than Deformable DETR even with only 10% encoder tokens on the COCO dataset. Albeit only the encoder tokens are sparsified, the total computation cost decreases by 38% and the frames per second (FPS) increases by 42% compared to Deformable DETR. Code is available at https://github.com/kakaobrain/sparse-detr

Byungseok Roh, Wuhyun Shin, Ildoo Kim et al.

Self-supervised learning has been widely used to obtain transferrable representations from unlabeled images. Especially, recent contrastive learning methods have shown impressive performances on downstream image classification tasks. While these contrastive methods mainly focus on generating invariant global representations at the image-level under semantic-preserving transformations, they are prone to overlook spatial consistency of local representations and therefore have a limitation in pretraining for localization tasks such as object detection and instance segmentation. Moreover, aggressively cropped views used in existing contrastive methods can minimize representation distances between the semantically different regions of a single image. In this paper, we propose a spatially consistent representation learning algorithm (SCRL) for multi-object and location-specific tasks. In particular, we devise a novel self-supervised objective that tries to produce coherent spatial representations of a randomly cropped local region according to geometric translations and zooming operations. On various downstream localization tasks with benchmark datasets, the proposed SCRL shows significant performance improvements over the image-level supervised pretraining as well as the state-of-the-art self-supervised learning methods. Code is available at https://github.com/kakaobrain/scrl

Sunghun Kang, Junbum Cha, Jonghwan Mun et al.

Open-vocabulary object detection (OVOD) has recently gained significant attention as a crucial step toward achieving human-like visual intelligence. Existing OVOD methods extend target vocabulary from pre-defined categories to open-world by transferring knowledge of arbitrary concepts from vision-language pre-training models to the detectors. While previous methods have shown remarkable successes, they suffer from indirect supervision or limited transferable concepts. In this paper, we propose a simple yet effective method to directly learn region-text alignment for arbitrary concepts. Specifically, the proposed method aims to learn arbitrary image-to-text mapping for pseudo-labeling of arbitrary concepts, named Pseudo-Labeling for Arbitrary Concepts (PLAC). The proposed method shows competitive performance on the standard OVOD benchmark for noun concepts and a large improvement on referring expression comprehension benchmark for arbitrary concepts.

Daejin Jo, Jeeyoung Yun, Byungseok Roh et al.

With the rapid progress of speech language models (SLMs), discrete speech tokens have emerged as a core interface between speech and text, enabling unified modeling across modalities. Recent speech tokenization approaches aim to isolate semantic information from low-level acoustics to better align with language models. In particular, previous methods use SSL teachers such as HuBERT to extract semantic representations, which are then distilled into a semantic quantizer to suppress acoustic redundancy as well as capture content-related latent structures. However, they still produce speech token sequences significantly longer than their textual counterparts, creating challenges for efficient speech-language modeling. Reducing the frame rate is a natural solution, but standard techniques, such as rigid average pooling across frames, can distort or dilute the semantic structure required for effective LM alignment. To address this, we propose LM-SPT, a speech tokenization method that introduces a novel semantic distillation. Instead of directly matching teacher and student features via pooling, we reconstruct speech solely from semantic tokens and minimize the discrepancy between the encoded representations of the original and reconstructed waveforms, obtained from a frozen automatic speech recognition (ASR) encoder. This indirect yet data-driven supervision enables the tokenizer to learn discrete units that are more semantically aligned with language models. LM-SPT further incorporates architectural improvements to the encoder and decoder for speech tokenization, and supports multiple frame rates, including 25Hz, 12.5Hz, and 6.25Hz. Experimental results show that LM-SPT achieves superior reconstruction fidelity compared to baselines, and that SLMs trained with LM-SPT tokens achieve competitive performances on speech-to-text and consistently outperform baselines on text-to-speech tasks.

Byungseok Roh, Han-Cheol Cho, Myung-Ho Ju et al.

Recent advances in deep learning have enabled complex real-world use cases comprised of multiple vision tasks and detection tasks are being shifted to the edge side as a pre-processing step of the entire workload. Since running a deep model on resource-constraint devices is challenging, techniques for efficient inference methods are demanded. In this paper, we present an objectness-aware object detection method to reduce computational cost by sparsifying activation values on background regions where target objects don't exist. Sparsified activation can be exploited to increase inference speed by software or hardware accelerated sparse convolution techniques. To accomplish this goal, we incorporate a light-weight objectness mask generation (OMG) network in front of an object detection (OD) network so that it can zero out unnecessary background areas of an input image before being fed into the OD network. In experiments, by switching background activation values to zero, the average number of zero values increases further from 36% to 68% on MobileNetV2-SSDLite even with ReLU activation while maintaining accuracy on MS-COCO. This result indicates that the total MAC including both OMG and OD networks can be reduced to 62% of the original OD model when only non-zero multiply-accumulate operations are considered. Moreover, we show a similar tendency in heavy networks (VGG and RetinaNet) and an additional dataset (PASCAL VOC).

Sanghoon Hong, Byungseok Roh, Kye-Hyeon Kim et al.

In object detection, reducing computational cost is as important as improving accuracy for most practical usages. This paper proposes a novel network structure, which is an order of magnitude lighter than other state-of-the-art networks while maintaining the accuracy. Based on the basic principle of more layers with less channels, this new deep neural network minimizes its redundancy by adopting recent innovations including C.ReLU and Inception structure. We also show that this network can be trained efficiently to achieve solid results on well-known object detection benchmarks: 84.9% and 84.2% mAP on VOC2007 and VOC2012 while the required compute is less than 10% of the recent ResNet-101.

Kye-Hyeon Kim, Sanghoon Hong, Byungseok Roh et al.

This paper presents how we can achieve the state-of-the-art accuracy in multi-category object detection task while minimizing the computational cost by adapting and combining recent technical innovations. Following the common pipeline of "CNN feature extraction + region proposal + RoI classification", we mainly redesign the feature extraction part, since region proposal part is not computationally expensive and classification part can be efficiently compressed with common techniques like truncated SVD. Our design principle is "less channels with more layers" and adoption of some building blocks including concatenated ReLU, Inception, and HyperNet. The designed network is deep and thin and trained with the help of batch normalization, residual connections, and learning rate scheduling based on plateau detection. We obtained solid results on well-known object detection benchmarks: 83.8% mAP (mean average precision) on VOC2007 and 82.5% mAP on VOC2012 (2nd place), while taking only 750ms/image on Intel i7-6700K CPU with a single core and 46ms/image on NVIDIA Titan X GPU. Theoretically, our network requires only 12.3% of the computational cost compared to ResNet-101, the winner on VOC2012.