Shallow Feature Based Dense Attention Network for Crowd Counting
Authors: Yunqi Miao, Zijia Lin, Guiguang Ding, Jungong Han11765-11772
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on three benchmark datasets clearly demonstrate the superiority of SDANet when dealing with different scenarios. |
| Researcher Affiliation | Collaboration | Yunqi Miao,1 Zijia Lin,2 Guiguang Ding,3 Jungong Han1 1University of Warwick, Coventry, UK 2Microsoft Research, Beijing, China 3Tsinghua University, Beijing, China |
| Pseudocode | No | The paper includes architectural diagrams (Figure 3, Figure 4) but does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | In the experiment, three crowd counting benchmark datasets, the UCF CC 50 dataset (Idrees et al. 2013), the World Expo 10 dataset (Zhang et al. 2015), and the Shanghai Tech dataset (Zhang et al. 2016), are used to evaluate the performance of SDANet. |
| Dataset Splits | Yes | On the UCF CC 50 dataset, we performed a 5-fold cross-validation to evaluate the proposed method as suggested by (Idrees et al. 2013). and World Expo 10 dataset (Zhang et al. 2015) ... is divided into a training set (3380 frames) and a test set (600 frames). |
| Hardware Specification | No | The paper discusses computational aspects like 'computation cost' but does not specify any hardware details (e.g., specific GPU or CPU models, memory) used for running experiments. |
| Software Dependencies | No | The paper states 'The implementation of SDANet is based on the Py Torch framework' but does not provide a version number for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | Adam (Kingma and Ba 2014) algorithm with the initial learning rate of 1e-4 is adopted to optimize the SDANet. and Following the previous work (Zhang et al. 2016), 9 patches, where each patch is 1/4 of the image size, are cropped from each image to generate the training set. and Additionally, images are further augmented by randomly horizontal flipping. and where M is the dimension of ˆ DC, and is set to 32 throughout all experiments. and Additionally, ϵ = 0.0001 is set to avoid the denominator being zero. and where α = 0.01 is the empirical weight for LC. and Loss = Latt + Lmap. |