Top-Down Feedback for Crowd Counting Convolutional Neural Network
Authors: Deepak Babu Sam, R. Venkatesh Babu
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate the performance of our model on all major crowd datasets and show the effectiveness of top-down feedback. Table 1 reports performance of TDF-CNN along with other models. It is seen that TDF-CNN outperforms all other models by a significant margin both in terms of MAE and MSE. Moreover, the number of parameters for TDF-CNN is less than all other models. This emphasizes effectiveness of top-down feedback in correcting density predictions. |
| Researcher Affiliation | Academia | Deepak Babu Sam, R. Venkatesh Babu Video Analytics Lab, Indian Institute of Science, Bangalore 560012, INDIA {deepaksam, venky}@iisc.ac.in |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any information about the availability of open-source code for the described methodology. |
| Open Datasets | Yes | The Shanghaitech crowd counting dataset, which consist of 1198 annotated images, is introduced by (Zhang et al. 2016). UCF CC 50 (Idrees et al. 2013) is a small dataset of 50 annotated crowd scenes. The World Expo 10 dataset (Zhang et al. 2015) contains 1132 video sequences captured with 108 surveillance cameras in Shanghai 2010 World Expo. |
| Dataset Splits | Yes | For Part A, 300 images are used for training and the rest 182 images for testing. Similarly, 400 images of Part B are for training and 316 for testing. Since there is no separate test set, 5-fold crossvalidation is performed for evaluation of our model (Idrees et al. 2013; Zhang et al. 2015; Boominathan, Kruthiventi, and Babu 2016; Zhang et al. 2016). |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU models, CPU types). |
| Software Dependencies | No | The paper mentions using CNNs and SGD but does not specify any software versions for libraries, frameworks, or programming languages (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The l2 distance between the predicted density map and ground truth is used as the loss to train the CNN regressor. Standard Stochastic Gradient Descent (SGD) algorithm is applied on the parameters Θ to optimize Ll2. Training is done with patches which have 1/4th size of the original image. 9 patches are cropped at different locations from every image to augment the data as in (Zhang et al. 2016). The values of the regularizer constants are chosen empirically. In all our experiments, these regularizers are fixed as λ = 10 2 and μ = 10 3. |