Mining Object Parts from CNNs via Active Question-Answering

Abstract

Given a convolutional neural network (CNN) that is pre-trained for object classification, this paper proposes to use active question-answering to semanticize neural patterns in conv-layers of the CNN and mine part concepts. For each part concept, we mine neural patterns in the pre-trained CNN, which are related to the target part, and use these patterns to construct an And-Or graph (AOG) to represent a four-layer semantic hierarchy of the part. As an interpretable model, the AOG associates different CNN units with different explicit object parts. We use an active human-computer communication to incrementally grow such an AOG on the pre-trained CNN as follows. We allow the computer to actively identify objects, whose neural patterns cannot be explained by the current AOG. Then, the computer asks human about the unexplained objects, and uses the answers to automatically discover certain CNN patterns corresponding to the missing knowledge. We incrementally grow the AOG to encode new knowledge discovered during the active-learning process. In experiments, our method exhibits high learning efficiency. Our method uses about 1/6-1/3 of the part annotations for training, but achieves similar or better part-localization performance than fast-RCNN methods.

Published in CVPR 2017

References in corpus (7)

• Learning Deep Features for Discriminative Localization
• Detect What You Can: Detecting and Representing Objects using Holistic Models and Body Parts
• Understanding Deep Image Representations by Inverting Them
• Unsupervised Object Discovery and Localization in the Wild: Part-based Matching with Bottom-up Region Proposals
• Part Detector Discovery in Deep Convolutional Neural Networks
• Understanding deep features with computer-generated imagery
• Cooperative Training of Descriptor and Generator Networks

Cited by in corpus (6)

• Interpreting CNN Knowledge via an Explanatory Graph
• Unsupervised Learning of Neural Networks to Explain Neural Networks
• Examining CNN Representations with respect to Dataset Bias
• A Taught-Obesrve-Ask (TOA) Method for Object Detection with Critical Supervision
• Mining Interpretable AOG Representations from Convolutional Networks via Active Question Answering
• Explanatory Graphs for CNNs