Secure vision: Integrated anti-spoofing and deep-fake detection system using knowledge distillation approach

Despite recent advances in deep learning, the rise of edge devices, and the exponential growth of Internet of Things (IoT) connected devices undermine the performance of deep learning models. It is clear that the future of computing is moving to edge devices. Autonomous vehicles and self-driving cars have leveraged the power of computer vision, especially object detection to navigate through traffic safely. Nevertheless, to be able to drive on all types of roads, these new vehicles have to be equipped with a road anomaly detection system, which strikes the need for small deep learning models of detecting road anomalies that can be deployed on these vehicles for safer driving experience. However, the current deep learning models are not practical on embedded devices due to the heavy resource requirements of the models, as such cannot be deployed on embedded devices. This paper proposes a theoretical approach to building a lightweight model from a cumbersome pothole detection model that is suitable on edge devices using knowledge distillation. It presents the theoretical approach of knowledge distillation, why it is a better technique of model compression compared to the rest. It shows that a cumbersome model can be made lightweight without sacrificing accuracy and with a reduced time complexity and faster training time.

In this work, we consider transferring the structure information from large networks to compact ones for dense prediction tasks in computer vision. Previous knowledge distillation strategies used for dense prediction tasks often directly borrow the distillation scheme for image classification and perform knowledge distillation for each pixel separately, leading to sub-optimal performance. Here we propose to distill structured knowledge from large networks to compact networks, taking into account the fact that dense prediction is a structured prediction problem. Specifically, we study two structured distillation schemes: i) pair-wise distillation that distills the pair-wise similarities by building a static graph; and ii) holistic distillation that uses adversarial training to distill holistic knowledge. The effectiveness of our knowledge distillation approaches is demonstrated by experiments on three dense prediction tasks: semantic segmentation, depth estimation and object detection. Code is available at https://git.io/StructKD.

Recognizing objects in low-resolution images is a challenging task due to the lack of informative details. Recent studies have shown that knowledge distillation approaches can effectively transfer knowledge from a high-resolution teacher model to a low-resolution student model by aligning cross-resolution representations. However, these approaches still face limitations in adapting to the situation where the recognized objects exhibit significant representation discrepancies between training and testing images. In this study, we propose a cross-resolution relational contrastive distillation approach to facilitate low-resolution object recognition. Our approach enables the student model to mimic the behavior of a well-trained teacher model which delivers high accuracy in identifying high-resolution objects. To extract sufficient knowledge, the student learning is supervised with contrastive relational distillation loss, which preserves the similarities in various relational structures in contrastive representation space. In this manner, the capability of recovering missing details of familiar low-resolution objects can be effectively enhanced, leading to a better knowledge transfer. Extensive experiments on low-resolution object classification and low-resolution face recognition clearly demonstrate the effectiveness and adaptability of our approach.

Fire accidents are life-threatening catastrophes leading to losses of life, financial damage, climate change, and ecological destruction. Promptly and efficiently detecting and extinguishing fires is essential to reduce the loss of lives and damage. This study uses drone, edge computing, and artificial intelligence (AI) techniques, presenting novel methods for real-time fire detection. This proposed work utilizes a comprehensive dataset of 7187 fire images and advanced deep learning models, e.g., Detection Transformer (DETR), Detectron2, You Only Look Once YOLOv8, and Autodistill-based knowledge distillation techniques to improve the model performance. The knowledge distillation approach has been implemented with the YOLOv8m (medium) as the teacher (base) model. The distilled (student) frameworks are developed employing the YOLOv8n (Nano) and DETR techniques. The YOLOv8n attains the best performance with 95.21% detection accuracy and 0.985 F1 score. A powerful hardware setup, including a Raspberry Pi 5 microcontroller, Pi camera module 3, and a DJI F450 custom-built drone, has been constructed. The distilled YOLOv8n model has been deployed in the proposed hardware setup for real-time fire identification. The YOLOv8n model achieves 89.23% accuracy and an approximate frame rate of 8 for the conducted live experiments. Integrating deep learning techniques with drone and edge devices demonstrates the proposed system’s effectiveness and potential for practical applications in fire hazard mitigation.

Deep learning models have gained prominence in AI research due to their exceptional feature extraction capabilities. In fingerprint-based indoor localization, 1D and 2D convolutional neural networks (CNNs) are frequently utilized for their superior performance and minimal hardware requirements. However, these models primarily extract spatial features and often exhibit degraded real-time performance due to signal inconsistencies caused by shadowing, fading, and multipath effects. To address this, incorporating temporal variations through 3D CNNs can enhance outcomes, albeit with increased computational costs. This study proposes a knowledge distillation (KD) approach using 3D CNNs for indoor localization, to bridge the computational and performance gaps. The proposed KD framework employs a complex teacher model (TM) to train two simpler student models A (SM-A) and B (SM-B). We employ both response-based and feature-based KD methods to optimize the student models’ weights. The proposed techniques are evaluated using the UJILIB dataset collected from two floors of the University of Jaume I library. To enhance model performance, data augmentation is applied, yielding substantial improvements in test accuracy of 10.56% for TM, 16.19% for SM-A, and 16.93% for SM-B. The application of KD results in accuracy improvements of 5.45% for SM-A and 3.35% for SM-B, consequently reducing their average positioning errors (APEs) from 1.070 m and 1.030 m to 0.840 m and 0.944 m, respectively. Compared to KD-based SM-A, TM shows only 3.30% higher positioning accuracy, but with a 65.14% increase in computational complexity. Notably, the KD-enhanced SM-A also achieves a significantly lower APE of 0.84 m compared to the 2D CNN of 1.72 m.

Existing unsupervised person re-identification (Re-ID) methods have achieved remarkable performance by adopting an alternate clustering-training manner. However, they still suffer from camera variation, which results in an inconsistent feature space and unreliable pseudo labels that severely degrade the performance. In this paper, we propose a cross-camera self-distillation (CCSD) framework for unsupervised person Re-ID to alleviate the effect of camera variation. Specifically, in the clustering phase, we propose a camera-aware cluster refinement mechanism, which first splits each cluster into multiple clusters according to the camera views, and then refines them into more compact clusters. In the training phase, we first obtain the similarity between the samples and the refined clusters from the same and different cameras, and then transfer the knowledge of similarity distribution from intra-camera to cross-camera. Since the intra-camera similarity is free from camera variation, our knowledge distillation approach is able to learn a more consistent feature space across cameras. Extensive experiments demonstrate the superiority of our proposed CCSD against the state-of-the-art approaches on unsupervised person Re-ID.

A Knowledge Distillation Approach with Time Series Prediction and Predictive Elastic Telescopic Algorithm in Computing Force Network

GID: Global information distillation for medical semantic segmentation

Deep learning models have been shown to be vulnerable to adversarial attacks. Adversarial attacks are imperceptible perturbations added to an image such that the deep learning model misclassifies the image with a high confidence. Existing adversarial defenses validate their performance using only the classification accuracy. However, classification accuracy by itself is not a reliable metric to determine if the resulting image is "adversarial-free". This is a foundational problem for online image recognition applications where the ground-truth of the incoming image is not known and hence we cannot compute the accuracy of the classifier or validate if the image is "adversarial-free" or not. This paper proposes a novel privacy preserving framework for defending Black box classifiers from adversarial attacks using an ensemble of iterative adversarial image purifiers whose performance is continuously validated in a loop using Bayesian uncertainties. The proposed approach can convert a single-step black box adversarial defense into an iterative defense and proposes three novel privacy preserving Knowledge Distillation (KD) approaches that use prior meta-information from various datasets to mimic the performance of the Black box classifier. Additionally, this paper proves the existence of an optimal distribution for the purified images that can reach a theoretical lower bound, beyond which the image can no longer be purified. Experimental results on six public benchmark datasets namely: 1) Fashion-MNIST, 2) CIFAR-10, 3) GTSRB, 4) MIO-TCD, 5) Tiny-ImageNet, and 6) MS-Celeb show that the proposed approach can consistently detect adversarial examples and purify or reject them against a variety of adversarial attacks.

Knowledge Distillation (KD) aims to distill the dark knowledge of a high-powered teacher network into a student network, which can improve the capacity of student network and has been successfully applied to semantic segmentation. However, the standard knowledge distillation approaches merely represent the supervisory signal of teacher network as the dark knowledge, while ignoring the impact of network architecture during distillation. In this paper, we found that the student network with a more similar architecture against the teacher network obtains more performance gain from distillation. Therefore, a more generalized paradigm for knowledge distillation is to distill both the soft-label and the structure of the teacher network. We propose a novel Structural Distillation (SD) method which introduces the structural similarity constraints into vanilla knowledge distillation. We leverage Neural Architecture Search technique to search optimal student structure for semantic segmentation from a well-designed search space, which mimics the given teacher both in terms of soft-label and network structure. Experiment results make clear that our proposed method outperforms both the NAS with conventional Knowledge Distillation and human-designed methods, and achieves sota performance on the Cityscapes dataset under various platform-aware latency constraints. Furthermore, the best architecture discovered on Cityscapes also transfers well to the PASCAL VOC2012 dataset.

Neural machine translation (NMT) offers a novel alternative formulation of translation that is potentially simpler than statistical approaches. However to reach competitive performance, NMT models need to be exceedingly large. In this paper we consider applying knowledge distillation approaches (Bucila et al., 2006; Hinton et al., 2015) that have proven successful for reducing the size of neural models in other domains to the problem of NMT. We demonstrate that standard knowledge distillation applied to word-level prediction can be effective for NMT, and also introduce two novel sequence-level versions of knowledge distillation that further improve performance, and somewhat surprisingly, seem to eliminate the need for beam search (even when applied on the original teacher model). Our best student model runs 10 times faster than its state-of-the-art teacher with little loss in performance. It is also significantly better than a baseline model trained without knowledge distillation: by 4.2/1.7 BLEU with greedy decoding/beam search. Applying weight pruning on top of knowledge distillation results in a student model that has 13 times fewer parameters than the original teacher model, with a decrease of 0.4 BLEU.

Open World Object Detection (OWOD) is a novel computer vision task with a considerable challenge, bridging the gap between classic object detection (OD) and real-world object detection. In addition to detecting and classifying seen/known objects, OWOD algorithms are expected to localize all potential unseen/unknown objects and incrementally learn them. The large pre-trained vision-language grounding models (VLM, e.g., GLIP) have rich knowledge about the open world, but are limited by text prompts and cannot localize indescribable objects. However, there are many detection scenarios in which pre-defined language descriptions are unavailable during inference. In this paper, we attempt to specialize the VLM model for OWOD tasks by distilling its open-world knowledge into a language-agnostic detector. Surprisingly, we observe that the simple knowledge distillation approach leads to unexpected performance for unknown object detection, even with a small amount of data. Unfortunately, knowledge distillation for unknown objects severely affects the learning of detectors with conventional structures, leading to catastrophic damage to the model's ability to learn about known objects. To alleviate these problems, we propose the down-weight training strategy for knowledge distillation from vision-language model to single visual modality one. Meanwhile, we propose the cascade decoupled decoders that decouple the learning of localization and recognition to reduce the impact of category interactions of known and unknown objects on the localization learning process. Ablation experiments demonstrate that both of them are effective in mitigating the impact of open-world knowledge distillation on the learning of known objects. Additionally, to alleviate the current lack of comprehensive benchmarks for evaluating the ability of the open-world detector to detect unknown objects in the open world, we refine the benchmark for evaluating the performance of unknown object detection by augmenting annotations for unknown objects which we name"IntensiveSet$\scriptstyle\spadesuit$♠". Comprehensive experiments performed on OWOD, MS-COCO, and our proposed benchmarks demonstrate the effectiveness of our methods.

The deep convolutional neural network has made significant progress in cloud detection. However, the compromise between having a compact model and high accuracy has always been a challenging task in cloud detection for large-scale remote sensing imagery. A promising method to tackle this problem is knowledge distillation, which usually lets the compact model mimic the cumbersome model’s output to get better generalization. However, vanilla knowledge distillation methods cannot properly distill the characteristics of clouds in remote sensing images. In this paper, we propose a novel self-attention knowledge distillation approach for compact and accurate cloud detection, named Bidirectional Self-Attention Distillation (Bi-SAD). Bi-SAD lets a model learn from itself without adding additional parameters or supervision. With bidirectional layer-wise features learning, the model can get a better representation of the cloud’s textural information and semantic information, so that the cloud’s boundaries become more detailed and the predictions become more reliable. Experiments on a dataset acquired by GaoFen-1 satellite show that our Bi-SAD has a great balance between compactness and accuracy, and outperforms vanilla distillation methods. Compared with state-of-the-art cloud detection models, the parameter size and FLOPs are reduced by 100 times and 400 times, respectively, with a small drop in accuracy.

One of the greatest engineering feats in history is the construction of tunnels, and the management of tunnel safety depends heavily on the detection of tunnel defects. However, the real-time, portability, and accuracy issues with the present tunnel defect detection technique still exist. The study improves the traditional defect detection technology based on the knowledge distillation algorithm, the depth pooling residual structure is designed in the teacher network to enhance the ability to extract target features. Next, the MobileNetv3 lightweight network is built into the student network to reduce the number and volume of model parameters. The lightweight model is then trained in terms of both features and outputs using a multidimensional knowledge distillation approach. By processing the tunnel radar detection photos, the dataset is created. The experimental findings demonstrate that the multidimensional knowledge distillation approach greatly increases the detection efficiency: the number of parameters is decreased by 81.4%, from 16.03 MB to 2.98 MB, while the accuracy is improved by 2.5%, from 83.4% to 85.9%.

Medical image classification: Knowledge transfer via residual U-Net and vision transformer-based teacher-student model with knowledge distillation