In-Shop Clothing Retrieval Dataset

In-Shop stands for In-shop Clothes Retrieval Benchmark, which can be used to evaluate the performance of clothing retrieval. This dataset is a large subset of DeepFashion and contains rich pose and scale variations.

The dataset also has a variety of large and rich annotations, including:

• 7,982 costumes;
• 52,712 in-shop clothing images, and approximately 200,000 cross-pose and scale pairs.

Each image is annotated with a bounding box, clothing type, and pose type.