There are tasks in Image processing,
- Image Classification
- Object Detection
- Semantic Segmentation
- Instance Segmentation
- Image Captioning
- Image Generation
- Image Denoising
- Pose Estimation
- Object Tracking
- 3D Reconstruction
- Scene Understanding
- Visual Question Answering
- Adversarial Attacks
Image classification
- assigning a label to an entire image (cat, dog, airplane)
convolution layers, pooling layers, activation layers
Class Activation Map (CAM)
- Using GAP (global average pooling) instead of the FC-layer
Object Detection
- Object Classification + Localization
object detection 모델은 객체의 위치(localization)를 찾고, class를 분류(classification)하는 작업을 진행
RCNN
Fast-RCNN
Faster-RCNN
DETR : Transformer Encoder - Decoder + Hungarian Matching
ReinaNet
Hard Negative Mining vs Focal Loss
Image Generation
Generator + Discriminator
Reference
AI researcher