Segment anything model (SAM) has demonstrated excellent generalization capabilities in common vision scenarios, yet lacking an understanding of specialized data. Although numerous works have focused on optimizing SAM for downstream tasks, these task-specific approaches usually limit the generalizability to other downstream tasks. In this paper, we aim to investigate the impact of the general vision modules on finetuning SAM and enable them to generalize across all downstream tasks. We propose a simple unified framework called SimAda for adapting SAM in underperformed scenes. Specifically, our framework abstracts the general modules of different methods into basic design elements, and we design four variants based on a shared theoretical framework. SimAda is simple yet effective, which removes all dataset-specific designs and focuses solely on general optimization, ensuring that SimAda can be applied to all SAM-based and even Transformer-based models. We conduct extensive experiments on nine datasets of six downstream tasks. The results demonstrate that SimAda significantly improves the performance of SAM on multiple downstream tasks and achieves state-of-the-art performance on most of them, without requiring task-specific designs. Code is available at: https://github.com/zongzi13545329/SimAda
Contour based scene text detection methods have rapidly developed recently, but still suffer from inaccurate frontend contour initialization, multi-stage error accumulation, or deficient local information aggregation. To tackle these limitations, we propose a novel arbitrary-shaped scene text detection framework named CT-Net by progressive contour regression with contour transformers. Specifically, we first employ a contour initialization module that generates coarse text contours without any post-processing. Then, we adopt contour refinement modules to adaptively refine text contours in an iterative manner, which are beneficial for context information capturing and progressive global contour deformation. Besides, we propose an adaptive training strategy to enable the contour transformers to learn more potential deformation paths, and introduce a re-score mechanism that can effectively suppress false positives. Extensive experiments are conducted on four challenging datasets, which demonstrate the accuracy and efficiency of our CT-Net over state-of-the-art methods. Particularly, CT-Net achieves F-measure of 86.1 at 11.2 frames per second (FPS) and F-measure of 87.8 at 10.1 FPS for CTW1500 and Total-Text datasets, respectively.
* This paper has been accepted by IEEE Transactions on Circuits and
Systems for Video Technology
The quality of point clouds is often limited by noise introduced during their capture process. Consequently, a fundamental 3D vision task is the removal of noise, known as point cloud filtering or denoising. State-of-the-art learning based methods focus on training neural networks to infer filtered displacements and directly shift noisy points onto the underlying clean surfaces. In high noise conditions, they iterate the filtering process. However, this iterative filtering is only done at test time and is less effective at ensuring points converge quickly onto the clean surfaces. We propose IterativePFN (iterative point cloud filtering network), which consists of multiple IterationModules that model the true iterative filtering process internally, within a single network. We train our IterativePFN network using a novel loss function that utilizes an adaptive ground truth target at each iteration to capture the relationship between intermediate filtering results during training. This ensures that the filtered results converge faster to the clean surfaces. Our method is able to obtain better performance compared to state-of-the-art methods. The source code can be found at: https://github.com/ddsediri/IterativePFN.
* This paper has been accepted to the IEEE/CVF CVPR Conference, 2023
Arbitrary-shaped scene text detection is a challenging task due to the variety of text changes in font, size, color, and orientation. Most existing regression based methods resort to regress the masks or contour points of text regions to model the text instances. However, regressing the complete masks requires high training complexity, and contour points are not sufficient to capture the details of highly curved texts. To tackle the above limitations, we propose a novel light-weight anchor-free text detection framework called TextDCT, which adopts the discrete cosine transform (DCT) to encode the text masks as compact vectors. Further, considering the imbalanced number of training samples among pyramid layers, we only employ a single-level head for top-down prediction. To model the multi-scale texts in a single-level head, we introduce a novel positive sampling strategy by treating the shrunk text region as positive samples, and design a feature awareness module (FAM) for spatial-awareness and scale-awareness by fusing rich contextual information and focusing on more significant features. Moreover, we propose a segmented non-maximum suppression (S-NMS) method that can filter low-quality mask regressions. Extensive experiments are conducted on four challenging datasets, which demonstrate our TextDCT obtains competitive performance on both accuracy and efficiency. Specifically, TextDCT achieves F-measure of 85.1 at 17.2 frames per second (FPS) and F-measure of 84.9 at 15.1 FPS for CTW1500 and Total-Text datasets, respectively.
* This paper has been accepted by IEEE Transactions on Multimedia
Abnormal event detection is a challenging task that requires effectively handling intricate features of appearance and motion. In this paper, we present an approach of detecting anomalies in videos by learning a novel LSTM based self-contained network on normal dense optical flow. Due to their sigmoid implementations, standard LSTM's forget gate is susceptible to overlooking and dismissing relevant content in long sequence tasks like abnormality detection. The forget gate mitigates participation of previous hidden state for computation of cell state prioritizing current input. In addition, the hyperbolic tangent activation of standard LSTMs sacrifices performance when a network gets deeper. To tackle these two limitations, we introduce a bi-gated, light LSTM cell by discarding the forget gate and introducing sigmoid activation. Specifically, the LSTM architecture we come up with fully sustains content from previous hidden state thereby enabling the trained model to be robust and make context-independent decision during evaluation. Removing the forget gate results in a simplified and undemanding LSTM cell with improved performance effectiveness and computational efficiency. Empirical evaluations show that the proposed bi-gated LSTM based network outperforms various LSTM based models verifying its effectiveness for abnormality detection and generalization tasks on CUHK Avenue and UCSD datasets.
* 16 pages, 7 figures, Computer Graphics International (CGI) 2020
Fine-grained facial expression manipulation is a challenging problem, as fine-grained expression details are difficult to be captured. Most existing expression manipulation methods resort to discrete expression labels, which mainly edit global expressions and ignore the manipulation of fine details. To tackle this limitation, we propose an end-to-end expression-guided generative adversarial network (EGGAN), which utilizes structured latent codes and continuous expression labels as input to generate images with expected expressions. Specifically, we adopt an adversarial autoencoder to map a source image into a structured latent space. Then, given the source latent code and the target expression label, we employ a conditional GAN to generate a new image with the target expression. Moreover, we introduce a perceptual loss and a multi-scale structural similarity loss to preserve identity and global shape during generation. Extensive experiments show that our method can manipulate fine-grained expressions, and generate continuous intermediate expressions between source and target expressions.
Abnormality detection is a challenging task due to the dependence on a specific context and the unconstrained variability of practical scenarios. In recent years, it has benefited from the powerful features learnt by deep neural networks, and handcrafted features specialized for abnormality detectors. However, these approaches with large complexity still have limitations in handling long term sequential data (e.g., videos), and their learnt features do not thoroughly capture useful information. Recurrent Neural Networks (RNNs) have been shown to be capable of robustly dealing with temporal data in long term sequences. In this paper, we propose a novel version of Gated Recurrent Unit (GRU), called Single Tunnelled GRU for abnormality detection. Particularly, the Single Tunnelled GRU discards the heavy weighted reset gate from GRU cells that overlooks the importance of past content by only favouring current input to obtain an optimized single gated cell model. Moreover, we substitute the hyperbolic tangent activation in standard GRUs with sigmoid activation, as the former suffers from performance loss in deeper networks. Empirical results show that our proposed optimized GRU model outperforms standard GRU and Long Short Term Memory (LSTM) networks on most metrics for detection and generalization tasks on CUHK Avenue and UCSD datasets. The model is also computationally efficient with reduced training and testing time over standard RNNs.
* H. Fanta Z. Shao and L. Ma Journal of Information Sciences 524
(2020) 15-32 * 14 pages, 11 figures, 13 tables, this paper is accepted on Journal of
Facial action unit (AU) detection and face alignment are two highly correlated tasks, since facial landmarks can provide precise AU locations to facilitate the extraction of meaningful local features for AU detection. However, most existing AU detection works handle the two tasks independently by treating face alignment as a preprocessing, and often use landmarks to predefine a fixed region or attention for each AU. In this paper, we propose a novel end-to-end deep learning framework for joint AU detection and face alignment, which has not been explored before. In particular, multi-scale shared feature is learned firstly, and high-level feature of face alignment is fed into AU detection. Moreover, to extract precise local features, we propose an adaptive attention learning module to refine the attention map of each AU adaptively. Finally, the assembled local features are integrated with face alignment feature and global feature for AU detection. Extensive experiments demonstrate that our framework (i) significantly outperforms the state-of-the-art AU detection methods on the challenging BP4D, DISFA, GFT and BP4D+ benchmarks, (ii) can adaptively capture the irregular region of each AU, (iii) achieves competitive performance for face alignment, and (iv) also works well under partial occlusions and non-frontal poses. The code for our method is available at https://github.com/ZhiwenShao/PyTorch-JAANet.
* arXiv admin note: text overlap with arXiv:1803.05588
Automatic facial action unit (AU) recognition has attracted great attention but still remains a challenging task, as subtle changes of local facial muscles are difficult to thoroughly capture. Most existing AU recognition approaches leverage geometry information in a straightforward 2D or 3D manner, which either ignore 3D manifold information or suffer from high computational costs. In this paper, we propose a novel geodesic guided convolution (GeoConv) for AU recognition by embedding 3D manifold information into 2D convolutions. Specifically, the kernel of GeoConv is weighted by our introduced geodesic weights, which are negatively correlated to geodesic distances on a coarsely reconstructed 3D face model. Moreover, based on GeoConv, we further develop an end-to-end trainable framework named GeoCNN for AU recognition. Extensive experiments on BP4D and DISFA benchmarks show that our approach significantly outperforms the state-of-the-art AU recognition methods.