A Deep Learning Framework for Joint Image Restoration and Recognition (original) (raw)
Related papers
Image Reconstruction Using Deep Neural Networks Models
International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2022
Image restoration is the process of restoring the original image. It can be challenging to eliminate image blur in a variety of contexts, including photography, radar imaging, and the removal of motion blur brought on by camera shaking. Image noise is unintentional signal that enters an image from a sensor, such as a thermal or electrical signal or an external factor like rain or snow. The image degradation may be caused by transmission noise, object motion, resolution restrictions, coding artefacts, camera shake, or a combination of these factors. In order to distinguish between HF and LF artefacts, image decomposition is employed to divide the deformed image into a texture layer and a structure layer (Low Frequency LF Component) The current approach utilises the frequency characteristics of various forms of artefacts through a configurable deep neural network structure. Therefore, by changing the architecture, the same method may be applied to a number of picture restoration tasks. A quality enhancement network that uses residual and recursive learning is suggested for decreasing the artefacts with comparable frequency characteristics. Residual learning is used to enhance performance and speed up the training process. Recursive learning is used to both improve performance and drastically cut down on the amount of training parameters. This Project aims to build systems for reconstructing the old images from under sampled one and mismatched Pixels to form a proper image to increase its visible quality and its pixels quality by using a Deep Neural network Models and it can improve the integration of various feature representations from many photos. Result Shows Improved Training accuracy of 92%.When compared to the two-frame designs now in use, the multi-frame architecture will be used which prevents repetitive computations caused by multiple inferences when aligning multiple images
Image restoration using deep learning
2016
We propose a new image restoration method that reduces noise and blur in degraded images. In contrast to many state of the art methods, our method does not rely on intensive iterative approaches, instead it uses a pre-trained convolutional neural network.
Restoration of artwork using deep neural networks
Evolving Systems, 2019
Paintings and other similar work of art represent an important part of our heritage and contemporary culture. However, due to the nature of materials used in these works, they are prone to damage and degradation over a period of time. Some of damages to these works may include torn canvases, smudges, exposure to elements etc. This necessitates the need for restoration of artworks. The restoration process is very time consuming and is a delicate task making it prone to human error. The virtual restoration of digitized artworks can be very helpful in this process. In this paper, we have proposed a method based on deep neural networks for virtual restoration of the digitized artworks. The paper presents a hybrid model which employs automatic mask generation based on Mask R-CNN and image inpainting using U-Net architecture with partial convolutions and automatic mask update. The proposed approach is evaluated qualitatively as well as quantitatively. The qualitative evaluation of the approach is done by engaging three domain art experts. On the other hand, quantitative validation of the proposed method is done using dataset of images having artificially created irregular holes by employing mean square error (MSE) and structural similarity index (SSIM) metrics. The results obtained show that the proposed approach is quite effective in virtual restoration of the digitized artworks.
Can fully convolutional networks perform well for general image restoration problems?
2017 Fifteenth IAPR International Conference on Machine Vision Applications (MVA), 2017
We present a fully convolutional network(FCN) based approach for color image restoration. FCNs have recently shown remarkable performance for high-level vision problem like semantic segmentation. In this paper, we investigate if FCN models can show promising performance for low-level problems like image restoration as well. We propose a fully convolutional model, that learns a direct end-to-end mapping between the corrupted images as input and the desired clean images as output. Our proposed method takes inspiration from domain transformation techniques but presents a data-driven task specific approach where filters for novel basis projection, task dependent coefficient alterations, and image reconstruction are represented as convolutional networks. Experimental results show that our FCN model outperforms traditional sparse coding based methods and demonstrates competitive performance compared to the state-of-the-art methods for image denoising. We further show that our proposed model can solve the difficult problem of blind image inpainting and can produce reconstructed images of impressive visual quality.
Knowledge-driven training of deep models for better reconstruction and recognition
Ph. D. Thesis of IISc, 2019
This thesis aims to efficiently solve many interesting and challenging problems by incorporating appropriate image processing techniques in a deep learning framework. We have proposed, implemented, and tested efficient and effective solutions for tasks such as image reconstruction and recognition. We have shown that the performance of any deep architecture can be improved at three different levels: (i) input, (ii) architecture, and (iii) the objective or loss function. While developing algorithms, our focus has been on designing architectures that can optimally utilize the advantages of our exploration at all these levels. In the first part, we propose different techniques to enhance the quality of low-resolution document images (particularly binary) for better human readability and OCR performance. We start with a comprehensive study on low-resolution images, showing that the performance of OCR can be improved by increasing the resolution of the document images. Next, we have improved the quality of low-resolution, down-sampled document images and real low-resolution document images. We have achieved significant enhancement in the quality of the reconstructed images, whereby humans find the reconstructed high-resolution images easy to read and the OCR recognition accuracy is also significantly improved. \textbf{(exploration of the input and architecture )} In the second part, we have proposed different techniques to improve the quality of low-resolution natural images. Here we have fused multiple interpolations in a deep network to obtain better reconstruction. This idea of fusing multiple interpolations can be applied to various computer vision and image processing tasks. This suggests that traditional algorithms can be combined in a deep framework to obtain better reconstruction (exploration of the input and architecture). In the third part, we have proposed mean square Canny error (MSCE) as a ``new loss function" that improves the performance of any existing deep architecture (super-resolution or denoising) that earlier used mean square error (MSE) as a loss function. Many a work in the literature use 'mean square error' in various super-resolution and denoising tasks. Our main goal in proposing this loss function is that it can improve all these existing algorithms (super-resolution or denoising) that use the mean square as a loss function without incurring additional costs during inference (exploration of the objective function). The fourth part of the thesis addresses practical applications of deep learning to some computer vision tasks. This part suggests that increasing the width and depth of a deep network is not always a better approach in the process of obtaining an optimal model (lightweight or less complex). We have shown that feeding the gradient and/or the Laplacian of the input image can improve the performance of facial emotion classifiers by a good margin, without incurring additional overhead during inference. This allows us to find a lightweight and computationally efficient model, without compromising the classification accuracies. In another task of real-time, artistic style transfer, we have proposed techniques to make it computationally more efficient, without much decrease in the perceptual quality of the reconstructed artistic images. We have proposed using depth-wise separable convolution (DepSep) in place of convolution and nearest neighbor (NN) interpolation in place of transposed convolution. We have also explored the concatenation of nearest neighbour and bilinear (Bil) interpolations in place of transposed convolution. The stylized images from the modified architectures are perceptually similar in quality to those from the original architecture. The decrease in the computational complexity of our architectures is validated by the decrease in the testing time by 26.1%, 39.1%, and 57.1%, respectively, for DepSep, DepSep-NN-Bil, and DepSep-NN variants (explorations of the input, architecture and the objective).
Image Restoration and Enhancement Using Deep Learning
International Journal of Engineering Applied Sciences and Technology
During the process of image acquisition, sometimes images are degraded because of various reasons like low resolution of camera, motion blur, noise etc. This paper presents the work associated with the Image Restoration & Enhancement techniques. The process of recovering degraded image is known as Image Restoration. Image restoration includes denoising image, image inpainting, etc. Here we proposed Convolution Neural Network (CNN) with Median Filter for removing noise, Region filling Exemplar Based Inpainting Algorithm for image inpainting. Image enhancement is one amongst the problem in image processing. Haze, low lighting etc. are the various problems in images. The aim of Image enhancement is to process an image such that result is more suitable than original image for specific application. Here for haze removal we implement dark channel prior algorithm and for lightning low-light image we proposed functions. Image enhancement improves the appearance of the image.
IJIRIS:: AM Publications,India, 2024
Image restoration is an integral component of computer vision that tries to restore pictures that have been deteriorated or corrupted to their original or enhanced condition. In this study, we look into the wide picture restoration techn models. There perform quite well, particularly when i rely on handcrafted filters restricts their adaptation to more complicated forms of been revolutionized by deep learning, which is led by co learning sophisticated representations of visual data. It is because of this that CNNs are able to deal with a wide variety of degradations, such as noise, blurring, artifacts, and missing data. Ge GANs, are continually pushing the limits of what is possible by utilizing adversarial training to accomplish spectacular outcomes in the areas of to overcome: Understanding limited interpretability of the the training of successful models may be quite computationally rigorous. make navigation revolutionize image processing and analysis, ultimately contributing to advancements across a wide range of scientific and technological domains. This can be concentrating on the promising research directions that are currently being pursued.
VEHICLE DAMAGED DETECTION USING DEEP LEARNING
International Research Journal of Modernization in Engineering Technology and Science, 2022
Object Detection based Vehicle Damage Detection system can potentially save insurance companies Million. In this paper we propose a Vehicle Damage Detection System based on YOLO v4. We created a dataset of damaged vehicles and annotated regions consisting of Dents, Shattered Glass, damaged tail lights and scratches. We then pre-processed the images inline with the requirements of YOLO v4 and trained the model to achieve an mAP@50 of 81.20. The system proposed in this paper surpasses all previously proposed methods and gives promising results.
2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021
Vehicle re-identification has the objective of finding a specific vehicle among different vehicle crops captured by multiple cameras placed at multiple intersections. Among the different difficulties, high intra-class variability and high inter-class similarity can be highlighted. Moreover, the resolution of the images can be different, which also means a challenge in the re-identification task. Intending to face these problems, we use as baseline our previous work based on obtaining different deep learning features and ensembling them to get a single, stable and robust feature vector. It also includes post-processing techniques that explode all the information provided by the CityFlowV2-ReID dataset, including a re-ranking step. Then, in this paper, several newly included improvements are described. Background and orientation similarity matrices are added to the system to reduce bias towards these characteristics. Furthermore, we take into account the camera labels to penalize the gallery images that share camera with the query image. Additionally, to improve the training step, a synthetic dataset is added to the original one.
CAR DAMAGE DETECTION USING DEEP LEARNING
IRJET, 2022
Image-based vehicle insurance processing is a key industry with a lot of possibilities for automation. We look at the subject of car damage categorization in this paper, where certain damages are classed as minor and others as serious. In several of the categories, fine-granularity is feasible. We go into the bowels of knowledge. We'll use based ways for this. We attempted to train a CNN straight at first. However, it does not work well with data due to the small number of tagged samples. The impact of pre-training in the domain is next examined, followed by fine-tuning. Finally, we put ensemble and transfer learning to the test. According to research, transfer learning outperforms domain-specific fine-tuning. We have a high level of commitment.