Research map / Video / Action recognition
FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras
Shanghang Zhang, Guanhang Wu, Joao P. Costeira, Jose M. F. Moura
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hypercolumn feature combination to atrous FCN (ablation) | FCN-A · WebCamT (downtown cameras) | MAE: 3.07 -> 2.04 | Improved |
| add atrous convolution to hypercolumn FCN (ablation) | FCN-H · WebCamT (downtown cameras) | MAE: 2.48 -> 2.04 | Improved |
| add hypercolumn feature combination to atrous FCN (ablation) | FCN-A · WebCamT (parkway cameras) | MAE: 2.75 -> 2.04 | Improved |
| add atrous convolution to hypercolumn FCN (ablation) | FCN-H · WebCamT (parkway cameras) | MAE: 2.30 -> 2.04 | Improved |
| add direct-connected LSTM on top of FCN-HA (ablation) | FCN-HA · WebCamT (downtown cameras) | MAE: 2.04 -> 1.80 | Improved |
| add direct-connected LSTM on top of FCN-HA (ablation) | FCN-HA · WebCamT (parkway cameras) | MAE: 2.04 -> 1.82 | Improved |
| use residual connection of FCN and LSTM instead of direct connection (ablation) | FCN-dLSTM · WebCamT (downtown cameras) | MAE: 1.80 -> 1.53 | Improved |
| use residual connection of FCN and LSTM instead of direct connection (ablation) | FCN-dLSTM · WebCamT (parkway cameras) | MAE: 1.82 -> 1.63 | Improved |
| use residual connection instead of direct connection for training speed (ablation) | FCN-dLSTM · WebCamT | convergence speed | Improved |
| add residual LSTM on FCN-HA (ablation) | FCN-HA · TRANCOS | MAE: 4.21 -> 4.38 | Hurt |
| add residual LSTM on FCN-HA (ablation) | FCN-HA · UCSD | MAE: 1.65 -> 1.54 | Improved |
| add residual LSTM on FCN-HA (ablation) | FCN-HA · UCSD | MSE: 3.37 -> 3.02 | Improved |
Closest papers by meaning
- Understanding Traffic Density From Large-Scale Web Camera Data (CVPR 2017 · similarity 0.91)
- A Large Contextual Dataset for Classification, Detection and Counting of Cars with Deep Learning (ECCV 2016 · similarity 0.87)
- Spatiotemporal Modeling for Crowd Counting in Videos (ICCV 2017 · similarity 0.86)
- DAVE: A Unified Framework for Fast Vehicle Detection and Annotation (ECCV 2016 · similarity 0.86)
- From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer (ICCV 2019 · similarity 0.85)
- Rethinking Spatial Invariance of Convolutional Networks for Object Counting (CVPR 2022 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
