Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userCompressing Visual-linguistic Model via Knowledge Distillation
104
0
0.0
(
0
)
Added by
Zhiyuan Fang
Publication date
2021
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Despite exciting progress in pre-training for visual-linguistic (VL) representations, very few aspire to a small VL model. In this paper, we study knowledge distillation (KD) to effectively compress a transformer-based large VL model into a small VL model. The major challenge arises from the inconsistent regional visual tokens extracted from different detectors of Teacher and Student, resulting in the misalignment of hidden representations and attention distributions. To address the problem, we retrain and adapt the Teacher by using the same region proposals from Students detector while the features are from Teachers own object detector. With aligned network inputs, the adapted Teacher is capable of transferring the knowledge through the intermediate representations. Specifically, we use the mean square error loss to mimic the attention distribution inside the transformer block and present a token-wise noise contrastive loss to align the hidden state by contrasting with negative representations stored in a sample queue. To this end, we show that our proposed distillation significantly improves the performance of small VL models on image captioning and visual question answering tasks. It reaches 120.8 in CIDEr score on COCO captioning, an improvement of 5.1 over its non-distilled counterpart; and an accuracy of 69.8 on VQA 2.0, a 0.8 gain from the baseline. Our extensive experiments and ablations confirm the effectiveness of VL distillation in both pre-training and fine-tuning stages.
rate research
Read More
Visual relationship detection aims to reason over relationships among salient objects in images, which has drawn increasing attention over the past few years. Inspired by human reasoning mechanisms, it is believed that external visual commonsense knowledge is beneficial for reasoning visual relationships of objects in images, which is however rarely considered in existing methods. In this paper, we propose a novel approach named Relational Visual-Linguistic Bidirectional Encoder Representations from Transformers (RVL-BERT), which performs relational reasoning with both visual and language commonsense knowledge learned via self-supervised pre-training with multimodal representations. RVL-BERT also uses an effective spatial module and a novel mask attention module to explicitly capture spatial information among the objects. Moreover, our model decouples object detection from visual relationship recognition by taking in object names directly, enabling it to be used on top of any object detection system. We show through quantitative and qualitative experiments that, with the transferred knowledge and novel modules, RVL-BERT achieves competitive results on two challenging visual relationship detection datasets. The source code is available at https://github.com/coldmanck/RVL-BERT.
Zero-shot image classification has made promising progress by training the aligned image and text encoders. The goal of this work is to advance zero-shot object detection, which aims to detect novel objects without bounding box nor mask annotations. We propose ViLD, a training method via Vision and Language knowledge Distillation. We distill the knowledge from a pre-trained zero-shot image classification model (e.g., CLIP) into a two-stage detector (e.g., Mask R-CNN). Our method aligns the region embeddings in the detector to the text and image embeddings inferred by the pre-trained model. We use the text embeddings as the detection classifier, obtained by feeding category names into the pre-trained text encoder. We then minimize the distance between the region embeddings and image embeddings, obtained by feeding region proposals into the pre-trained image encoder. During inference, we include text embeddings of novel categories into the detection classifier for zero-shot detection. We benchmark the performance on LVIS dataset by holding out all rare categories as novel categories. ViLD obtains 16.1 mask AP$_r$ with a Mask R-CNN (ResNet-50 FPN) for zero-shot detection, outperforming the supervised counterpart by 3.8. The model can directly transfer to other datasets, achieving 72.2 AP$_{50}$, 36.6 AP and 11.8 AP on PASCAL VOC, COCO and Objects365, respectively.
Existing online knowledge distillation approaches either adopt the student with the best performance or construct an ensemble model for better holistic performance. However, the former strategy ignores other students information, while the latter increases the computational complexity. In this paper, we propose a novel method for online knowledge distillation, termed FFSD, which comprises two key components: Feature Fusion and Self-Distillation, towards solving the above problems in a unified framework. Different from previous works, where all students are treated equally, the proposed FFSD splits them into a student leader and a common student set. Then, the feature fusion module converts the concatenation of feature maps from all common students into a fused feature map. The fused representation is used to assist the learning of the student leader. To enable the student leader to absorb more diverse information, we design an enhancement strategy to increase the diversity among students. Besides, a self-distillation module is adopted to convert the feature map of deeper layers into a shallower one. Then, the shallower layers are encouraged to mimic the transformed feature maps of the deeper layers, which helps the students to generalize better. After training, we simply adopt the student leader, which achieves superior performance, over the common students, without increasing the storage or inference cost. Extensive experiments on CIFAR-100 and ImageNet demonstrate the superiority of our FFSD over existing works. The code is available at https://github.com/SJLeo/FFSD.
Multi-label image classification is a fundamental but challenging task towards general visual understanding. Existing methods found the region-level cues (e.g., features from RoIs) can facilitate multi-label classification. Nevertheless, such methods usually require laborious object-level annotations (i.e., object labels and bounding boxes) for effective learning of the object-level visual features. In this paper, we propose a novel and efficient deep framework to boost multi-label classification by distilling knowledge from weakly-supervised detection task without bounding box annotations. Specifically, given the image-level annotations, (1) we first develop a weakly-supervised detection (WSD) model, and then (2) construct an end-to-end multi-label image classification framework augmented by a knowledge distillation module that guides the classification model by the WSD model according to the class-level predictions for the whole image and the object-level visual features for object RoIs. The WSD model is the teacher model and the classification model is the student model. After this cross-task knowledge distillation, the performance of the classification model is significantly improved and the efficiency is maintained since the WSD model can be safely discarded in the test phase. Extensive experiments on two large-scale datasets (MS-COCO and NUS-WIDE) show that our framework achieves superior performances over the state-of-the-art methods on both performance and efficiency.
This paper addresses the problem of model compression via knowledge distillation. To this end, we propose a new knowledge distillation method based on transferring feature statistics, specifically the channel-wise mean and variance, from the teacher to the student. Our method goes beyond the standard way of enforcing the mean and variance of the student to be similar to those of the teacher through an $L_2$ loss, which we found it to be of limited effectiveness. Specifically, we propose a new loss based on adaptive instance normalization to effectively transfer the feature statistics. The main idea is to transfer the learned statistics back to the teacher via adaptive instance normalization (conditioned on the student) and let the teacher network evaluate via a loss whether the statistics learned by the student are reliably transferred. We show that our distillation method outperforms other state-of-the-art distillation methods over a large set of experimental settings including different (a) network architectures, (b) teacher-student capacities, (c) datasets, and (d) domains.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
