Evaluation Task and Evaluation Data Introduction

Vision Domain Evaluation (CV)

Vision domain (CV) mainly evaluates the three major capabilities of models in downstream tasks: 1) perceptual capability, including local perception and temporal perception; 2) analytical capability, including global analysis, local analysis, and temporal analysis; 3) understanding capability, including analogy, induction, and reasoning.

Currently, it includes the following evaluation tasks:

• Depth Estimation：including evaluation datasets of NYUv2, SUN RGB-D, KITTI, DDAD, etc.
• Image Classification：including evaluation datasets of ImageNet, Place365, etc.
• Image Retrieval：including evaluation datasets of SOP, iNaturalist, etc.
• Semantic Segmentation：including evaluation datasets of ADE20K, COCO-Stuff, Cityscapes, etc.
• Semi-Supervised Image Classification：including evaluation datasets of ImageNet, Place365, etc.
• Small Sample Image Classification： including evaluation datasets of ImageNet, Place365, Stanford Cars, CUB-200-2011, FGVC-Aircraft, Food-101, DTD, etc.

Click the task name to view the task details. The task details include task introduction, evaluation data introduction, and evaluation metrics introduction.