buct-vision / paper-notes

Paper Notes in Computer Vision and Deep Learning(Machine Learning)

CV-DL Recommandations

Recommanded resources in Computer Vision and Deep Learning including advanced paper and issue-solutions in experiments.

Contents

• CV-DL Recommandations
  • Contents
    • Backbone Network
    • Image Recognition
    • Super Resolution
    • Object Detection
    • Semantic Segmentation
    • Image Caption
    • Generative Adversarial Networks
    • Attention Mechanism
    • Natural Language Processing Related
    • Dataset and Contest
      • 1. Dataset
      • 2. Contest
    • Training Techniques

Backbone Network

Image Recognition

Super Resolution

Object Detection

Semantic Segmentation

• Vortex Pooling: Improving Context Representation in Semantic Segmentation(2018.4) [pdf]
• Pyramid Attention Network for Semantic Segmentation(2018.5) [pdf]

Image Caption

• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention(2015) [pdf] [code_tensorflow] [code_PyTorch]
• Image Captioning with Semantic Attention(2016) [pdf] [code]
• Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering(2017) [pdf][code]
• Convolutional Image Captioning(2017) [pdf] [code]
• CNN+CNN: Convolutional Decoders for Image Captioning (2018) [pdf]

Generative Adversarial Networks

• Video-to-Video Synthesis(2018) [pdf] [code_PyTorch]
• Diverse Image-to-Image Translation via Disentangled Representations(2018.8) [pdf] [code_PyTorch] (Notes: maybe suitable for unpaired MR-CT synthesis for human body)

Attention Mechanism

• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention(2015) [pdf] [code_TensorFlow] [code_PyTorch]
• Image Captioning with Semantic Attention(2016) [pdf] [code]
• Attention Is All You Need(2017) [pdf] [code_PyTorch] [code_TensorFlow]
• Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering(2017) [pdf] [code]
• Attention U-Net:Learning Where to Look for the Pancreas(2018) [pdf] [code]
• Self-Attention Generative Adversarial Networks(2018.5) [pdf] [code_PyTorch] (Notes: 将自我注意机制引入到GAN的生成模型中，对于图像的纹理和几何上的联系提供全局的注意使得生成的图像更加的合理)
• Pervasive Attention: 2D Convolutional Neural Networks for Sequence-to-Sequence Prediction(2018) [pdf] [code]

Natural Language Processing Related

• Pervasive Attention: 2D Convolutional Neural Networks for Sequence-to-Sequence Prediction(2018) [pdf] [code]

Dataset and Contest

1. Dataset

• PASCAL VOC 2012
• Cityscapes
• Grand Challenges in Biomedical Image Analysis
• NYUv2
• PASCAL-Context
• SUNRGBD
• ADE20k

2. Contest

• AI Challenger

Training Techniques

• An overview of gradient descent optimization algorithms