Research map / Detection and segmentation / CNN features
A Large Contextual Dataset for Classification, Detection and Counting of Cars with Deep Learning
T. Nathan Mundhenk, Goran Konjevod, Wesam A. Sakla, Kofi Boakye
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 residual projection shortcut to Inception layers (ResCeption) | GoogLeNet/Inception · COWC patch classification | patch accuracy: 99.12% -> 99.14% | Improved |
| use Inception/ResCeption instead of AlexNet | classification network · COWC patch classification | patch accuracy: 97.62% -> 99.14% | Improved |
| train on non-Utah data and test on Utah | ResCeption · COWC patch classification | patch accuracy: 99.14% -> 98.89% | Mixed |
| increase visible context margin up to 192x192 pixels (ablation) | GoogLeNet · COWC patch classification | patch accuracy | Improved |
| add residual shortcut (ResCeption) over Inception | counting network · COWC patch counting | MAE: 0.257 -> 0.255 | Improved |
| double the depth of ResCeption with one error output (Taller o1) | ResCeption · COWC patch counting | exactly correct count: 80.34% -> 81.07% | Improved |
| use Inception/ResCeption instead of AlexNet | counting network · COWC patch counting | exactly correct count: 67.97% -> 80.35% | Improved |
| add three error outputs to the taller ResCeption (ablation) | ResCeption Taller · COWC scene counting | total error: 1.19% -> 0.46% | Improved |
| average counts over four offset strides | ResCeption Taller o3 · COWC scene counting | MAE: 6.14% -> 5.15% | Improved |
| average counts over four offset strides | ResCeption Taller o3 · COWC scene counting | total error: 0.46% -> 0.20% | Improved |
| use AlexNet instead of larger networks for speed | AlexNet · COWC scene counting | time per 2048x2048 scene (s): 0.087 vs 0.773 | Mixed |
| raise softmax output to the power of 16 | ResCeption heat map · COWC detection | heat map gradient | No clear effect |
Closest papers by meaning
- FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras (ICCV 2017 · similarity 0.87)
- Drone-Based Object Counting by Spatially Regularized Regional Proposal Network (ICCV 2017 · similarity 0.85)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.85)
- Counting With Focus for Free (ICCV 2019 · similarity 0.85)
- MoVie: Revisiting Modulated Convolutions for Visual Counting and Beyond (ICLR 2021 · similarity 0.85)
- Learning to Count without Annotations (CVPR 2024 · similarity 0.83)
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.
