Automated diagnosis with artificial intelligence has emerged as a promising area in the realm of medical imaging, while the interpretability of the introduced deep neural networks still remains an urgent concern. Although contemporary works, such as XProtoNet and MProtoNet, has sought to design interpretable prediction models for the issue, the localization precision of their resulting attribution maps can be further improved. To this end, we propose a Multi-scale Attentive Prototypical part Network, termed MAProtoNet, to provide more precise maps for attribution. Specifically, we introduce a concise multi-scale module to merge attentive features from quadruplet attention layers, and produces attribution maps. The proposed quadruplet attention layers can enhance the existing online class activation mapping loss via capturing interactions between the spatial and channel dimension, while the multi-scale module then fuses both fine-grained and coarse-grained information for precise maps generation. We also apply a novel multi-scale mapping loss for supervision on the proposed multi-scale module. Compared to existing interpretable prototypical part networks in medical imaging, MAProtoNet can achieve state-of-the-art performance in localization on brain tumor segmentation (BraTS) datasets, resulting in approximately 4% overall improvement on activation precision score (with a best score of 85.8%), without using additional annotated labels of segmentation. Our code will be released in https://github.com/TUAT-Novice/maprotonet.
In this paper, we pursue a novel 3D AIGC setting: generating 3D content from IDEAs. The definition of an IDEA is the composition of multimodal inputs including text, image, and 3D models. To our knowledge, this challenging and appealing 3D AIGC setting has not been studied before. We propose the novel framework called Idea-2-3D to achieve this goal, which consists of three agents based upon large multimodel models (LMMs) and several existing algorithmic tools for them to invoke. Specifically, these three LMM-based agents are prompted to do the jobs of prompt generation, model selection and feedback reflection. They work in a cycle that involves both mutual collaboration and criticism. Note that this cycle is done in a fully automatic manner, without any human intervention. The framework then outputs a text prompt to generate 3D models that well align with input IDEAs. We show impressive 3D AIGC results that are beyond any previous methods can achieve. For quantitative comparisons, we construct caption-based baselines using a whole bunch of state-of-the-art 3D AIGC models and demonstrate Idea-2-3D out-performs significantly. In 94.2% of cases, Idea-2-3D meets users' requirements, marking a degree of match between IDEA and 3D models that is 2.3 times higher than baselines. Moreover, in 93.5% of the cases, users agreed that Idea-2-3D was better than baselines. Codes, data and models will made publicly available.
Deep neural network-based image compression (NIC) has achieved excellent performance, but NIC method models have been shown to be susceptible to backdoor attacks. Adversarial training has been validated in image compression models as a common method to enhance model robustness. However, the improvement effect of adversarial training on model robustness is limited. In this paper, we propose a prior knowledge-guided adversarial training framework for image compression models. Specifically, first, we propose a gradient regularization constraint for training robust teacher models. Subsequently, we design a knowledge distillation based strategy to generate a priori knowledge from the teacher model to the student model for guiding adversarial training. Experimental results show that our method improves the reconstruction quality by about 9dB when the Kodak dataset is elected as the backdoor attack object for psnr attack. Compared with Ma2023, our method has a 5dB higher PSNR output at high bitrate points.
We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models.
In this paper, we present a new data-efficient voxel-based self-supervised learning method for event cameras. Our pre-training overcomes the limitations of previous methods, which either sacrifice temporal information by converting event sequences into 2D images for utilizing pre-trained image models or directly employ paired image data for knowledge distillation to enhance the learning of event streams. In order to make our pre-training data-efficient, we first design a semantic-uniform masking method to address the learning imbalance caused by the varying reconstruction difficulties of different regions in non-uniform data when using random masking. In addition, we ease the traditional hybrid masked modeling process by explicitly decomposing it into two branches, namely local spatio-temporal reconstruction and global semantic reconstruction to encourage the encoder to capture local correlations and global semantics, respectively. This decomposition allows our selfsupervised learning method to converge faster with minimal pre-training data. Compared to previous approaches, our self-supervised learning method does not rely on paired RGB images, yet enables simultaneous exploration of spatial and temporal cues in multiple scales. It exhibits excellent generalization performance and demonstrates significant improvements across various tasks with fewer parameters and lower computational costs.
Object-goal navigation is a challenging task that requires guiding an agent to specific objects based on first-person visual observations. The ability of agent to comprehend its surroundings plays a crucial role in achieving successful object finding. However, existing knowledge-graph-based navigators often rely on discrete categorical one-hot vectors and vote counting strategy to construct graph representation of the scenes, which results in misalignment with visual images. To provide more accurate and coherent scene descriptions and address this misalignment issue, we propose the Aligning Knowledge Graph with Visual Perception (AKGVP) method for object-goal navigation. Technically, our approach introduces continuous modeling of the hierarchical scene architecture and leverages visual-language pre-training to align natural language description with visual perception. The integration of a continuous knowledge graph architecture and multimodal feature alignment empowers the navigator with a remarkable zero-shot navigation capability. We extensively evaluate our method using the AI2-THOR simulator and conduct a series of experiments to demonstrate the effectiveness and efficiency of our navigator. Code available: https://github.com/nuoxu/AKGVP.
Tensor network structure search (TN-SS), aiming at searching for suitable tensor network (TN) structures in representing high-dimensional problems, largely promotes the efficacy of TN in various machine learning applications. Nonetheless, finding a satisfactory TN structure using existing algorithms remains challenging. To develop more effective algorithms and avoid the human labor-intensive development process, we explore the knowledge embedded in large language models (LLMs) for the automatic design of TN-SS algorithms. Our approach, dubbed GPTN-SS, leverages an elaborate crafting LLM-based prompting system that operates in an evolutionary-like manner. The experimental results, derived from real-world data, demonstrate that GPTN-SS can effectively leverage the insights gained from existing methods to develop novel TN-SS algorithms that achieve a better balance between exploration and exploitation. These algorithms exhibit superior performance in searching the high-quality TN structures for natural image compression and model parameters compression while also demonstrating generalizability in their performance.
A modern deep neural network (DNN) for image classification tasks typically consists of two parts: a backbone for feature extraction, and a head for feature encoding and class predication. We observe that the head structures of mainstream DNNs adopt a similar feature encoding pipeline, exploiting global feature dependencies while disregarding local ones. In this paper, we revisit the feature encoding problem, and propose Non-glObal Attentive Head (NOAH) that relies on a new form of dot-product attention called pairwise object category attention (POCA), efficiently exploiting spatially dense category-specific attentions to augment classification performance. NOAH introduces a neat combination of feature split, transform and merge operations to learn POCAs at local to global scales. As a drop-in design, NOAH can be easily used to replace existing heads of various types of DNNs, improving classification performance while maintaining similar model efficiency. We validate the effectiveness of NOAH on ImageNet classification benchmark with 25 DNN architectures spanning convolutional neural networks, vision transformers and multi-layer perceptrons. In general, NOAH is able to significantly improve the performance of lightweight DNNs, e.g., showing 3.14\%|5.3\%|1.9\% top-1 accuracy improvement to MobileNetV2 (0.5x)|Deit-Tiny (0.5x)|gMLP-Tiny (0.5x). NOAH also generalizes well when applied to medium-size and large-size DNNs. We further show that NOAH retains its efficacy on other popular multi-class and multi-label image classification benchmarks as well as in different training regimes, e.g., showing 3.6\%|1.1\% mAP improvement to large ResNet101|ViT-Large on MS-COCO dataset. Project page: https://github.com/OSVAI/NOAH.
The deep neural networks are known to be vulnerable to well-designed adversarial attacks. The most successful defense technique based on adversarial training (AT) can achieve optimal robustness against particular attacks but cannot generalize well to unseen attacks. Another effective defense technique based on adversarial purification (AP) can enhance generalization but cannot achieve optimal robustness. Meanwhile, both methods share one common limitation on the degraded standard accuracy. To mitigate these issues, we propose a novel framework called Adversarial Training on Purification (AToP), which comprises two components: perturbation destruction by random transforms (RT) and purifier model fine-tuned (FT) by adversarial loss. RT is essential to avoid overlearning to known attacks resulting in the robustness generalization to unseen attacks and FT is essential for the improvement of robustness. To evaluate our method in an efficient and scalable way, we conduct extensive experiments on CIFAR-10, CIFAR-100, and ImageNette to demonstrate that our method achieves state-of-the-art results and exhibits generalization ability against unseen attacks.
Surgical tool segmentation and action recognition are fundamental building blocks in many computer-assisted intervention applications, ranging from surgical skills assessment to decision support systems. Nowadays, learning-based action recognition and segmentation approaches outperform classical methods, relying, however, on large, annotated datasets. Furthermore, action recognition and tool segmentation algorithms are often trained and make predictions in isolation from each other, without exploiting potential cross-task relationships. With the EndoVis 2022 SAR-RARP50 challenge, we release the first multimodal, publicly available, in-vivo, dataset for surgical action recognition and semantic instrumentation segmentation, containing 50 suturing video segments of Robotic Assisted Radical Prostatectomy (RARP). The aim of the challenge is twofold. First, to enable researchers to leverage the scale of the provided dataset and develop robust and highly accurate single-task action recognition and tool segmentation approaches in the surgical domain. Second, to further explore the potential of multitask-based learning approaches and determine their comparative advantage against their single-task counterparts. A total of 12 teams participated in the challenge, contributing 7 action recognition methods, 9 instrument segmentation techniques, and 4 multitask approaches that integrated both action recognition and instrument segmentation.