Evidential deep learning (EDL) models, based on Subjective Logic, introduce a principled and computationally efficient way to make deterministic neural networks uncertainty-aware. The resulting evidential models can quantify fine-grained uncertainty using learned evidence. However, the Subjective-Logic framework constrains evidence to be non-negative, requiring specific activation functions whose geometric properties can induce activation-dependent learning-freeze behavior-a regime where gradients become extremely small for samples mapped into low-evidence regions. We theoretically characterize this behavior and analyze how different evidential activations influence learning dynamics. Building on this analysis, we design a general family of activation functions and corresponding evidential regularizers that provide an alternative pathway for consistent evidence updates across activation regimes. Extensive experiments on four benchmark classification problems (MNIST, CIFAR-10, CIFAR-100, and Tiny-ImageNet), two few-shot classification problems, and blind face restoration problem empirically validate the developed theory and demonstrate the effectiveness of the proposed generalized regularized evidential models.

This paper proposes a method that enhances the compression performance of the current model under development for the upcoming MPEG standard on Feature Compression for Machines (FCM) [1]. By truncating low-activation feature channels and signaling these truncations in the bitstream, the method reduces bitrate while preserving task accuracy. Experimental results show an average BD-rate reduction of 10.59% across datasets and tasks, demonstrating gains in bitrate efficiency.

Advancements in text-to-image generative AI with large multimodal models are spreading into the field of image compression, creating high-quality representation of images at extremely low bit rates. This work introduces novel components to the existing multimodal image semantic compression (MISC) approach, enhancing the quality of the generated images in terms of PSNR and perceptual metrics. The new components include semantic segmentation guidance for the generative decoder, as well as content-adaptive diffusion, which controls the number of diffusion steps based on image characteristics. The results show that our newly introduced methods significantly improve the baseline MISC model while also decreasing the complexity. As a result, both the encoding and decoding time are reduced by more than 36%. Moreover, the proposed compression framework outperforms mainstream codecs in terms of perceptual similarity and quality. The code and visual examples are available.1

The split-inference paradigm divides an artificial intelligence (AI) model into two parts. This necessitates the transfer of intermediate feature data between the two halves. Here, effective compression of the feature data becomes vital. In this paper, we employ Z-score normalization to efficiently recover the compressed feature data at the decoder side. To examine the efficacy of our method, the proposed method is integrated into the latest Feature Coding for Machines (FCM) codec standard under development by the Moving Picture Experts Group (MPEG). Our method supersedes the existing scaling method used by the current standard under development. It both reduces the overhead bits and improves the end-task accuracy. To further reduce the overhead in certain circumstances, we also propose a simplified method. Experiments show that using our proposed method shows 17.09% reduction in bitrate on average across different tasks and up to 65.69% for object tracking without sacrificing the task accuracy.

Web refresh crawling is the problem of keeping a cache of web pages fresh, that is, having the most recent copy available when a page is requested, given a limited bandwidth available to the crawler. Under the assumption that the change and request events, resp., to each web page follow independent Poisson processes, the optimal scheduling policy was derived by Azar et al. 2018. In this paper, we study an extension of this problem where side information indicating content changes, such as various types of web pings, for example, signals from sitemaps, content delivery networks, etc., is available. Incorporating such side information into the crawling policy is challenging, because (i) the signals can be noisy with false positive events and with missing change events; and (ii) the crawler should achieve a fair performance over web pages regardless of the quality of the side information, which might differ from web page to web page. We propose a scalable crawling algorithm which (i) uses the noisy side information in an optimal way under mild assumptions; (ii) can be deployed without heavy centralized computation; (iii) is able to crawl web pages at a constant total rate without spikes in the total bandwidth usage over any time interval, and automatically adapt to the new optimal solution when the total bandwidth changes without centralized computation. Experiments clearly demonstrate the versatility of our approach.

Achieving successful variable bitrate compression with computationally simple algorithms from a single end-to-end learned image or video compression model remains a challenge. Many approaches have been proposed, including conditional auto-encoders, channel-adaptive gains for the latent tensor or uniformly quantizing all elements of the latent tensor. This paper follows the traditional approach to vary a single quantization step size to perform uniform quantization of all latent tensor elements. However, three modifications are proposed to improve the variable rate compression performance. First, multi objective optimization is used for (post) training. Second, a quantization-reconstruction offset is introduced into the quantization operation. Third, variable rate quantization is used also for the hyper latent. All these modifications can be made on a pre-trained single-rate compression model by performing post training. The algorithms are implemented into three well-known image compression models and the achieved variable rate compression results indicate negligible or minimal compression performance loss compared to training multiple models. (Codes will be shared at https://github.com/InterDigitalInc/CompressAI)

As an increasing amount of image and video content will be analyzed by machines, there is demand for a new codec paradigm that is capable of compressing visual input primarily for the purpose of computer vision inference, while secondarily supporting input reconstruction. In this work, we propose a learned compression architecture that can be used to build such a codec. We introduce a novel variational formulation that explicitly takes feature data relevant to the desired inference task as input at the encoder side. As such, our learned scalable image codec encodes and transmits two disentangled latent representations for object detection and input reconstruction. We note that compared to relevant benchmarks, our proposed scheme yields a more compact latent representation that is specialized for the inference task. Our experiments show that our proposed system achieves a bit rate savings of 40.6% on the primary object detection task compared to the current state-of-the-art, albeit with some degradation in performance for the secondary input reconstruction task.

Occluded person Re-identification is a challenging task which aims to find or distinguish a specific person when the human body is occluded by obstacles, other persons or oneself. Some recent state-of-the-art works adopting a transformer and/or pose-guided methods have improved the feature representation and performances, but are still in trouble with both weak representation and heavy structure. In this paper, we suggest the novel methods of transformer-based Re-identification for the occluded person as follows. First, in data augmentation, instead of deleting an arbitrary area, only a part containing the keypoint features of a person is deleted for effective learning in occlusion. Second, we suggest a unique hierarchical patch and feature based cross attention combining the reliability-enhanced heatmaps and the output of the transformer intermediate layer, which can more effectively pay attention to the non-occluded human region. Third, we propose an entire-partial loss function with non-similarity grouping for more robust feature representation. We compare our approach and various existing methods for the mAP and Rank-1 performances on the Occluded-Duke, Occluded-ReID, Market-1501 and DukeMTMC datasets. Experimental results demonstrate that our proposed model not only outperforms the existing methods, but also has the smallest scale among state-of-the-art methods.

Video content is watched not only by humans, but increasingly also by machines. For example, machine learning models analyze surveillance video for security and traffic moni-toring, search through YouTube videos for inappropriate content, and so on. In this paper, we propose a scalable video coding framework that supports machine vision (specifically, object detection) through its base layer bitstream and human vision via its enhancement layer bitstream. The proposed framework includes components from both conventional and Deep Neural Network (DNN)-based video coding. The results show that on object detection, the proposed framework achieves 13–19% bit savings compared to state-of-the-art video codecs, while remaining competitive in terms of MS-SSIM on the human vision task.

Spatial frequency analysis and transforms serve a central role in most engineered image and video lossy codecs, but are rarely employed in neural network (NN)-based approaches. We propose a novel NN-based image coding framework that utilizes forward wavelet transforms to decompose the input signal by spatial frequency. Our encoder generates separate bitstreams for each latent representation of low and high frequencies. This enables our decoder to selectively decode bitstreams in a quality-scalable manner. Hence, the decoder can produce an enhanced image by using an enhancement bitstream in addition to the base bitstream. Furthermore, our method is able to enhance only a specific region of interest (ROI) by using a corresponding part of the enhancement latent representation. Our experiments demonstrate that the proposed method shows competitive rate-distortion performance compared to several non-scalable image codecs. We also showcase the effectiveness of our two-level quality scalability, as well as its practicality in ROI quality enhancement.

At present, and increasingly so in the future, much of the captured visual content will not be seen by humans. Instead, it will be used for automated machine vision analytics and may require occasional human viewing. Examples of such applications include traffic monitoring, visual surveillance, autonomous navigation, and industrial machine vision. To address such requirements, we develop an end-to-end learned image codec whose latent space is designed to support scalability from simpler to more complicated tasks. The simplest task is assigned to a subset of the latent space (the base layer), while more complicated tasks make use of additional subsets of the latent space, i.e., both the base and enhancement layer(s). For the experiments, we establish a 2-layer and a 3-layer model, each of which offers input reconstruction for human vision, plus machine vision task(s), and compare them with relevant benchmarks. The experiments show that our scalable codecs offer 37%-80% bitrate savings on machine vision tasks compared to best alternatives, while being comparable to state-of-the-art image codecs in terms of input reconstruction.

We investigate latent-space scalability for multi-task collaborative intelligence, where one of the tasks is object detection and the other is input reconstruction. In our proposed approach, part of the latent space can be selectively decoded to support object detection while the remainder can be decoded when input reconstruction is needed. Such an approach allows reduced computational resources when only object detection is required, and this can be achieved without reconstructing input pixels. By varying the scaling factors of various terms in the training loss function, the system can be trained to achieve various trade-offs between object detection accuracy and input reconstruction quality. Experiments are conducted to demonstrate the adjustable system performance on the two tasks compared to the relevant benchmarks.

We propose a neural network model to estimate the current frame from two reference frames, using affine transformation and adaptive spatially-varying filters. The estimated affine transformation allows for using shorter filters compared to existing approaches for deep frame prediction. The predicted frame is used as a reference for coding the current frame. Since the proposed model is available at both encoder and decoder, there is no need to code or transmit motion information for the predicted frame. By making use of dilated convolutions and reduced filter length, our model is significantly smaller, yet more accurate, than any of the neural networks in prior works on this topic. Two versions of the proposed model - one for unidirectional, and one for bi-directional prediction - are trained using a combination of Discrete Cosine Transform (DCT)-based ℓ1-loss with various transform sizes, multi-scale Mean Squared Error (MSE) loss, and an object context reconstruction loss. The trained models are integrated with the HEVC video coding pipeline. The experiments show that the proposed models achieve about 7.3%, 5.4%, and 4.2% bit savings for the luminance component on average in the Low delay P, Low delay, and Random access configurations, respectively.

Recent AI applications such as Collaborative Intelligence with neural networks involve transferring deep feature tensors between various computing devices. This necessitates tensor compression in order to optimize the usage of bandwidth-constrained channels between devices. In this paper we present a prediction scheme called Back-and-Forth (BaF) prediction, developed for deep feature tensors, which allows us to dramatically reduce tensor size and improve its compressibility. Our experiments with a state-of-the-art object detector demonstrate that the proposed method allows us to significantly reduce the number of bits needed for compressing feature tensors extracted from deep within the model, with negligible degradation of the detection performance and without requiring any retraining of the network weights. We achieve a 62% and 75% reduction in tensor size while keeping the loss in accuracy of the network to less than 1% and 2%, respectively.

In collaborative intelligence applications, part of a deep neural network (DNN) is deployed on a relatively low-complexity device such as a mobile phone or edge device, and the remainder of the DNN is processed where more computing resources are available, such as in the cloud. This paper presents a novel lightweight compression technique designed specifically to code the activations of a split DNN layer, while having a low complexity suitable for edge devices and not requiring any retraining. We also present a modified entropy-constrained quantizer design algorithm optimized for clipped activations. When applied to popular object-detection and classification DNNs, we were able to compress the 32-bit floating point activations down to 0.6 to 0.8 bits, while keeping the loss in accuracy to less than 1%. When compared to HEVC, we found that the lightweight codec consistently provided better inference accuracy, by up to 1.3%. The performance and simplicity of this lightweight compression technique makes it an attractive option for coding a layer's activations in split neural networks for edge/cloud applications.

We propose a novel frame prediction method using a deep neural network (DNN), with the goal of improving the video coding efficiency. The proposed DNN makes use of decoded frames, at both the encoder and decoder to predict the textures of the current coding block. Unlike conventional inter-prediction, the proposed method does not require any motion information to be transferred between the encoder and the decoder. Still, both the uni-directional and bi-directional predictions are possible using the proposed DNN, which is enabled by the use of the temporal index channel, in addition to the color channels. In this paper, we developed a jointly trained DNN for both uni-directional and bi-directional predictions, as well as separate networks for uni-directional and bi-directional predictions, and compared the efficacy of both the approaches. The proposed DNNs were compared with the conventional motion-compensated prediction in the latest video coding standard, High Efficiency Video Coding (HEVC), in terms of the BD-bitrate. The experiments show that the proposed joint DNN (for both uni-directional and bi-directional predictions) reduces the luminance bitrate by about 4.4%, 2.4%, and 2.3% in the low delay $P$ , low delay, and random access configurations, respectively. In addition, using the separately trained DNNs brings further bit savings of about 0.3%-0.5%.

Finding faces in images is one of the most important tasks in computer vision, with applications in biometrics, surveillance, human-computer interaction, and other areas. In our earlier work, we demonstrated that it is possible to tell whether or not an image contains a face by only examining the HEVC syntax, without fully reconstructing the image. In the present work we move further in this direction by showing how to localize faces in HEVC-coded images, without full reconstruction. We also demonstrate the benefits that such approach can have in privacy-friendly face localization.

Image and video compression has traditionally been tailored to human vision. However, modern applications such as visual analytics and surveillance rely on computers “seeing” and analyzing the images before (or instead of) humans. For these applications, it is important to adjust compression to computer vision. In this paper we present a bit allocation and rate control strategy that is tailored to object detection. U sing the initial convolutional layers of a state-of-the-art object detector, we create an importance map that can guide bit allocation to areas that are important for object detection. The proposed method enables bit rate savings of 7% or more compared to default HEVC, at the equivalent object detection rate.

Recent studies have shown that the efficiency of deep neural networks in mobile applications can be significantly improved by distributing the computational workload between the mobile device and the cloud. This paradigm, termed collaborative intelligence, involves communicating feature data between the mobile and the cloud. The efficiency of such approach can be further improved by lossy compression of feature data, which has not been examined to date. In this work we focus on collaborative object detection and study the impact of both near-lossless and lossy compression of feature data on its accuracy. We also propose a strategy for improving the accuracy under lossy feature compression. Experiments indicate that using this strategy, the communication overhead can be reduced by up to 70% without sacrificing accuracy.

Collaborative intelligence is a new paradigm for efficient deployment of deep neural networks across the mobile-cloud infrastructure. By dividing the network between the mobile and the cloud, it is possible to distribute the computational workload such that the overall energy and/or latency of the system is minimized. However, this necessitates sending deep feature data from the mobile to the cloud in order to perform inference. In this work, we examine the differences between the deep feature data and natural image data, and propose a simple and effective near-lossless deep feature compressor. The proposed method achieves up to 5% bit rate reduction compared to HEVC-Intra and even more against other popular image codecs. Finally, we suggest an approach for reconstructing the input image from compressed deep features that could serve to supplement the inference performed by the deep model.

Corner-like features are important in computer vision problems such as object matching, tracking, recognition, and retrieval. Most corner detectors operate in the pixel domain, which means that they require image or video to be fully decoded and reconstructed before detection can start. In this paper we describe a method for generating corner proposals from compressed HEVC bitstreams without full decoding. Specifically, we utilize HEVC syntax and intra prediction directions to find the locations that are likely to contain corners. The proposed method is lightweight and can be applied to intra-coded frames or still images coded by HEVC. Experimental results illustrate that the proposed method is able to identify most regions where conventional pixel-domain corner detectors would find corners.