Abstract: Deployments of real-world object detection systems often experience a degradation in performance over time due to concept drift. Systems that leverage thermal cameras are especially susceptible because the respective thermal signatures of objects and their surroundings are highly sensitive to environmental changes. In this study, two types of weather-aware latent conditioning methods are investigated. The proposed method aims to guide two object detectors, (YOLOv5 and Deformable DETR) to become weather-aware. This is achieved by leveraging an auxiliary branch that predicts weather-related information while conditioning intermediate layers of the object detector. While the conditioning methods proposed do not directly improve the accuracy of baseline detectors, it can be observed that conditioned networks manage to extract a weather-related signal from the thermal images, thus resulting in a decreased miss rate at the cost of increased false positives. The extracted signal appears noisy and is thus challenging to regress accurately. This is most likely a result of the qualitative nature of the thermal sensor; thus, further work is needed to identify an ideal method for optimizing the conditioning branch, as well as to further improve the accuracy of the system.

Abstract: In modern e-commerce, item content features in various modalities offer accurate yet comprehensive information to recommender systems. The majority of previous work either focuses on learning effective item representation during modelling user-item interactions, or exploring item-item relationships by analysing multi-modal features. Those methods, however, fail to incorporate the collaborative item-user-item relationships into the multi-modal feature-based item structure. In this work, we propose a graph-based item structure enhancement method MM-GEF: Multi-Modal recommendation with Graph Early-Fusion, which effectively combines the latent item structure underlying multi-modal contents with the collaborative signals. Instead of processing the content feature in different modalities separately, we show that the early-fusion of multi-modal features provides significant improvement. MM-GEF learns refined item representations by injecting structural information obtained from both multi-modal and collaborative signals. Through extensive experiments on four publicly available datasets, we demonstrate systematical improvements of our method over state-of-the-art multi-modal recommendation methods.

Abstract: The goal of the challenge is to develop a test-time adaptation (TTA) method, which could adapt the model to gradually changing domains in video sequences for semantic segmentation task. It is based on a synthetic driving video dataset – SHIFT. The source model is trained on images taken during daytime in clear weather. Domain changes at test-time are mainly caused by varying weather conditions and times of day. The TTA methods are evaluated in each image sequence (video) separately, meaning the model is reset to the source model state before the next sequence. Images come one by one and a prediction has to be made at the arrival of each frame. Each sequence is composed of 401 images and starts with the source domain, then gradually drifts to a different one (changing weather or time of day) until the middle of the sequence. In the second half of the sequence, the domain gradually shifts back to the source one. Ground truth data is available only for the validation split of the SHIFT dataset, in which there are only six sequences that start and end with the source domain. We conduct an analysis specifically on those sequences. Ground truth data for test split, on which the developed TTA methods are evaluated for leader board ranking, are not publicly available.
The proposed solution secured a 3rd place in a challenge and received an innovation award. Contrary to the solutions that scored better, we did not use any external pretrained models or specialized data augmentations, to keep the solutions as general as possible. We have focused on analyzing the distributional shift and developing a method that could adapt to changing data dynamics and generalize across different scenarios.

Abstract: The field of visual document understanding has witnessed a rapid growth in emerging challenges and powerful multi-modal strategies. However, they rely on an extensive amount of document data to learn their pretext objectives in a ``pre-train-then-fine-tune'' paradigm and thus, suffer a significant performance drop in real-world online industrial settings. One major reason is the over-reliance on OCR engines to extract local positional information within a document page. Therefore, this hinders the model's generalizability, flexibility and robustness due to the lack of capturing global information within a document image. We introduce TransferDoc, a cross-modal transformer-based architecture pre-trained in a self-supervised fashion using three novel pretext objectives. TransferDoc learns richer semantic concepts by unifying language and visual representations, which enables the production of more transferable models. Besides, two novel downstream tasks have been introduced for a ``closer-to-real'' industrial evaluation scenario where TransferDoc outperforms other state-of-the-art approaches.

Abstract: Guided image processing techniques are widely used to extract meaningful information from a guiding image and facilitate the enhancement of the guided one. This paper specifically addresses the challenge of guided thermal image super-resolution, where a low-resolution thermal image is enhanced using a high-resolution visible spectrum image. We propose a new strategy that enhances outcomes from current guided super-resolution methods. This is achieved by transforming the initial guiding data into a representation resembling a thermal-like image, which is more closely in sync with the intended output. Experimental results with upscale factors of 8 and 16, demonstrate the outstanding performance of our approach in guided thermal image super-resolution obtained by mapping the original guiding information to a thermal-like image representation.

Abstract: Visual content is increasingly being processed by machines for various automated content analysis tasks instead of being consumed by humans. Despite the existence of several compression methods tailored for machine tasks, few consider real-world scenarios with multiple tasks. In this paper, we aim to address this gap by proposing a task-switchable pre-processor that optimizes input images specifically for machine consumption prior to encoding by an off-the-shelf codec designed for human consumption. The proposed task-switchable pre-processor adeptly maintains relevant semantic information based on the specific characteristics of different downstream tasks, while effectively suppressing irrelevant information to reduce bitrate. To enhance the processing of semantic information for diverse tasks, we leverage pre-extracted semantic features to modulate the pixel-to-pixel mapping within the pre-processor. By switching between different modulations, multiple tasks can be seamlessly incorporated into the system. Extensive experiments demonstrate the practicality and simplicity of our approach. It significantly reduces the number of parameters required for handling multiple tasks while still delivering impressive performance. Our method showcases the potential to achieve efficient and effective compression for machine vision tasks, supporting the evolving demands of real-world applications.

Abstract: In the field of continual learning, models are designed to learn tasks one after the other. While most research has centered on supervised continual learning, recent studies have highlighted the strengths of self-supervised continual representation learning. The improved transferability of representations built with self-supervised methods is often associated with the role played by the multi-layer perceptron projector. In this work, we depart from this observation and reexamine the role of supervision in continual representation learning. We reckon that additional information, such as human annotations, should not deteriorate the quality of representations. Our findings show that supervised models when enhanced with a multi-layer perceptron head, can outperform self-supervised models in continual representation learning.

Abstract: Sign Language Recognition (SLR) has garnered significant attention from researchers in recent years, particularly the intricate domain of Continuous Sign Language Recognition (CSLR), which presents heightened complexity compared to Isolated Sign Language Recognition (ISLR). One of the prominent challenges in CSLR pertains to accurately detecting the boundaries of isolated signs within a continuous video stream. Additionally, the reliance on handcrafted features in existing models poses a challenge to achieving optimal accuracy. To surmount these challenges, we propose a novel approach utilizing a Transformer-based model. Unlike traditional models, our approach focuses on enhancing accuracy while eliminating the need for handcrafted features. The Transformer model is employed for both ISLR and CSLR. The training process involves using isolated sign videos, where hand keypoint features extracted from the input video are enriched using the Transformer model. Subsequently, these enriched features are forwarded to the final classification layer. The trained model, coupled with a post-processing method, is then applied to detect isolated sign boundaries within continuous sign videos. The evaluation of our model is conducted on two distinct datasets, including both continuous signs and their corresponding isolated signs, demonstrates promising results.