GigaHumanDet: Exploring Full-Body Detection on Gigapixel-Level Images

Authors: Chenglong Liu, Haoran Wei, Jinze Yang, Jintao Liu, Wenxi Li, Yuchen Guo, Lu Fang

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on PANDA and STCrowd datasets show the superiority and strong applicability of our design.
Researcher Affiliation Collaboration 1University of Chinese Academy of Sciences 2BNRist, Tsinghua University 3MEGVII Technology 4Shanghai Jiao Tong University
Pseudocode No The paper describes the proposed method using descriptive text and mathematical equations, but it does not include any formal pseudocode blocks or algorithm environments.
Open Source Code No The paper does not provide any explicit statement or link indicating that the source code for their method is publicly available.
Open Datasets Yes PANDA. We evaluate the proposed Giga Human Det on the gigapixel-level human-centric dataset PANDA (Wang et al. 2020). STCrowd. STCrowd (Cong et al. 2022) is released recently and the total number of pedestrians is 219 K. There are 5263 and 2988 images with a size of 1280 × 720 in the training set and validation set, respectively.
Dataset Splits Yes PANDA. It provides 13 scenarios for training and 5 scenarios for testing. STCrowd. There are 5263 and 2988 images with a size of 1280 × 720 in the training set and validation set, respectively.
Hardware Specification Yes We train our model on four RTX 3090 GPUs with 4 × 24 GB RAM.
Software Dependencies No The paper mentions using specific backbones like Hourglass-104 and DLA-34, but it does not specify software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions).
Experiment Setup Yes The model with hourglass(Newell, Yang, and Deng 2016) or DLA-34(Yu et al. 2018) backbone is trained with a batch size of 43 for 40k iterations. The learning rate is set to 0.001, dropping 10 at the 30k iteration.