Abstract:The ability to correctly classify and retrieve apparel images has a variety of applications important to e-commerce, online advertising and internet search. In this work, we propose a robust framework for fine-grained apparel classification, in-shop and cross-domain retrieval which eliminates the requirement of rich annotations like bounding boxes and human-joints or clothing landmarks, and training of bounding box/ key-landmark detector for the same. Factors such as subtle appearance differences, variations in human poses, different shooting angles, apparel deformations, and self-occlusion add to the challenges in classification and retrieval of apparel items. Cross-domain retrieval is even harder due to the presence of large variation between online shopping images, usually taken in ideal lighting, pose, positive angle and clean background as compared with street photos captured by users in complicated conditions with poor lighting and cluttered scenes. Our framework uses compact bilinear CNN with tensor sketch algorithm to generate embeddings that capture local pairwise feature interactions in a translationally invariant manner. For apparel classification, we pass the feature embeddings through a softmax classifier, while, the in-shop and cross-domain retrieval pipelines use a triplet-loss based optimization approach, such that squared Euclidean distance between embeddings measures the dissimilarity between the images. Unlike previous works that relied on bounding box, key clothing landmarks or human joint detectors to assist the final deep classifier, proposed framework can be trained directly on the provided category labels or generated triplets for triplet loss optimization. Lastly, Experimental results on the DeepFashion fine-grained categorization, and in-shop and consumer-to-shop retrieval datasets provide a comparative analysis with previous work performed in the domain.

Cross-Scenario Clothing Retrieval And Fine-Grained Style Recognition

Fast Cross-Scenario Clothing Retrieval Based on Indexing Deep Features

Clothing retrieval combining hierarchical over-segmentation and cross-domain dictionary learning

Fashion Recommendation on Street Images.

Snap and Find: Deep Discrete Cross-domain Garment Image Retrieval

CSIR4G: an Effective and Efficient Cross-Scenario Image Retrieval Model for Glasses

Fine-grained Apparel Classification and Retrieval without rich annotations

Towards Better Understanding the Clothing Fashion Styles: A Multimodal Deep Learning Approach

MMFL-Net: Multi-scale and Multi-granularity Feature Learning for Cross-domain Fashion Retrieval

Two-Stage Method for Clothing Feature Detection

Robust and Automatic Clothing Reconstruction Based on a Single RGB Image.

Fine-grained convolutional neural network-based clothes recommendation method

Csir4g

Searching for Apparel Products from Images in the Wild

Multi-Task Curriculum Transfer Deep Learning of Clothing Attributes

Clothing Co-Parsing by Joint Image Segmentation and Labeling

BigFashion: A Large-Scale Dataset for Fine-Grained Attributes Recognition

Learning and Recognition of Clothing Genres From Full-Body Images

A Multi-Task Learning Model for Better Representation of Clothing Images

Interpretable Multimodal Retrieval for Fashion Products.

Garment Detectives: Discovering Clothes and Its Genre in Consumer Photos