Almost Unsupervised Learning for Dense Crowd Counting
Authors: Deepak Babu Sam, Neeraj N Sajjan, Himanshu Maurya, R. Venkatesh Babu8868-8875
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate the model performance on standard crowd counting datasets. Primarily two metrics are used by all supervised works on crowd counting. Count estimation accuracy of the model is inferred from the Mean Absolute Error or MAE metric. It is expressed as... Mean squared error or MSE is the second metric for model comparison. |
| Researcher Affiliation | Academia | Deepak Babu Sam, Neeraj N Sajjan, Himanshu Maurya, R. Venkatesh Babu Video Analytics Lab, Indian Institute of Science, Bangalore 560012, India deepaksam@iisc.ac.in, nnsajjan@gmail.com, himanshu.mib@gmail.com, venky@iisc.ac.in |
| Pseudocode | No | The paper does not include a dedicated section for pseudocode or an algorithm block. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing source code or a link to a code repository. |
| Open Datasets | Yes | The Shanghaitech dataset introduced by (Zhang et al. 2016) is the largest crowd counting dataset. Part A set of the dataset has 300 training images and 182 images for testing. |
| Dataset Splits | No | While the paper mentions that "The training is continued till loss Ll2 on the validation set stops improving," it does not provide specific details on the size or how the validation set was created (e.g., split percentages or counts). |
| Hardware Specification | No | The paper does not explicitly describe the hardware (e.g., specific GPU or CPU models, memory, or cloud instances) used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., |
| Experiment Setup | Yes | The GWTA cell size is chosen to be 32 32 and is subsequently halved after every pooling layer so that grid dimensions remain same across layers... The decoder De Conv1 is a transposed convolution with its weight tied with that of Conv1. Note that we do not have bias for the encoder and decoder, which we find to be empirically better... Parameters Θ are obtained by optimizing Ll2 with stochastic gradient descent (SGD)... In this work, we use a sigma of 8.0 for generating ground truth density maps... For a given test image, overlapping patches (10% overlap) are obtained and evaluated on the trained model. The density map predictions of the overlapping areas are averaged to obtain the final density map. |