DecAug: Augmenting HOI Detection via Decomposition

Authors: Hao-Shu Fang, Yichen Xie, Dian Shao, Yong-Lu Li, Cewu Lu1300-1308

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that our method brings up to 3.3 m AP and 1.6 m AP improvements on V-COCO and HICO-DET dataset for two advanced models. Specifically, interactions with fewer samples enjoy more notable improvement. Our method can be easily integrated into various HOI detection models with negligible extra computational consumption.
Researcher Affiliation Academia 1 Shanghai Jiao Tong University 2 The Chinese University of Hong Kong
Pseudocode No No pseudocode or algorithm block was found.
Open Source Code Yes Our code will be made publicly available.
Open Datasets Yes We evaluate our methods on two mainstream benchmarks: V-COCO (Gupta and Malik 2015) and HICO-DET (Chao et al. 2018).
Dataset Splits Yes V-COCO... It includes 10,346 images (2,533 for training, 2,867 for validating and 4,946 for testing)
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for running experiments were found. The paper only mentions "without burdening GPUs."
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions) were found.
Experiment Setup Yes Hyper-parameters We adopt stochastic gradient descent in training. All hyper-parameters strictly follow the original setting of our baseline models including iteration number, learning rate, weight decay, backbones and so on. Augmentation Pipeline During training, the proposed local and global augmentation strategies are incorporated simultaneously since they are complimentary. Each input image will be augmented with a probability of 0.5.