Growing Interpretable Part Graphs on ConvNets via Multi-Shot Learning

Authors: Quanshi Zhang, Ruiming Cao, Ying Nian Wu, Song-Chun Zhu

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our method is guided by a small number of part annotations, and it achieves superior performance (about 13% 107% improvement) in part center prediction on the PASCAL VOC and Image Net datasets.
Researcher Affiliation Academia Quanshi Zhang, Ruiming Cao, Ying Nian Wu, Song-Chun Zhu University of California, Los Angeles
Pseudocode No The paper describes the learning process and equations but does not include explicit pseudocode or algorithm blocks.
Open Source Code Yes 1Codes here: https://sites.google.com/site/cnnsemantics/
Open Datasets Yes We chose the 16-layer VGG network (VGG-16) (Simonyan and Zisserman 2015) that was pre-trained using the 1.3M images in the Image Net ILSVRC 2012 dataset (Deng et al. 2009) for object classification. ... We tested our method on three benchmark datasets: the PASCAL VOC Part Dataset (Chen et al. 2014), the CUB2002011 dataset (Wah et al. 2011), and the ILSVRC 2013 DET dataset (Deng et al. 2009).
Dataset Splits No In Experiments, we annotated 3 12 boxes for each part to build the AOG, and we used the rest images in the dataset as testing images. No explicit validation split is mentioned.
Hardware Specification No The paper does not provide specific details about the hardware used, such as GPU or CPU models.
Software Dependencies No The paper mentions using VGG-16 but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes We chose the 16-layer VGG network (VGG-16)... We chose the last 9 (from the 5-th to the 13th) conv-layers as valid conv-layers... Exp. 1, three-shot AOG construction: ... used a total of three annotations... Exp. 2, AOG construction with more annotations: ... annotated four parts in four different object images.