Research map / Architectures and efficiency / CNN design
Learning to Steer by Mimicking Features from Heterogeneous Auxiliary Networks
Yuenan Hou, Zheng Ma, Chunxiao Liu, Chen Change Loy
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 multi-layer feature mimicking from PSPNet and FlowNet2 (ablation) | 3D ResNet+LSTM · Udacity | MAE: 1.7147 -> 1.6236 | Improved |
| Add multi-layer feature mimicking from PSPNet and FlowNet2 (ablation) | 3D ResNet+LSTM · Comma.ai | MAE: 0.7989 -> 0.7048 | Improved |
| Add multi-layer feature mimicking from PSPNet and FlowNet2 | 3D ResNet + LSTM · BDD100K | Accuracy: 83.69% -> 85.03% | Improved |
| Add only high-level feature mimicking from PSPNet and FlowNet2 (ablation) | 3D ResNet+LSTM · Udacity | RMSE: 2.4899 -> 2.4013 | Improved |
| Add only middle-level feature mimicking from PSPNet and FlowNet2 (ablation) | 3D ResNet+LSTM · Udacity | RMSE: 2.4899 -> 2.4659 | Improved |
| Add only low-level feature mimicking from PSPNet and FlowNet2 (ablation) | 3D ResNet+LSTM · Udacity | RMSE: 2.4899 -> 2.4521 | Improved |
| Add PSPNet feature mimicking at all three levels (ablation) | 3D ResNet+LSTM · Udacity | MAE: 1.7147 -> 1.6653 | Improved |
| Add FlowNet2 feature mimicking at all three levels (ablation) | 3D ResNet+LSTM · Udacity | MAE: 1.7147 -> 1.6573 | Improved |
| Add feature mimicking from PSPNet and FlowNet2 instead of using neither (ablation) | FM-Net · Udacity | MAE: 1.7147 -> 1.6653 | Improved |
| Add Cityscape pre-training only (ablation) | FM-Net · Udacity | MAE: 1.7147 -> 1.7125 | No clear effect |
| Use two-stage training, introducing mimicking after steering and multitask training (ablation) | FM-Net · Steering angle prediction | prediction performance | Improved |
| Add a temporal LSTM module (ablation) | 3D ResNet · Udacity | MAE: 1.9167 -> 1.7147 | Improved |
Closest papers by meaning
- Gradient-Free Adversarial Training Against Image Corruption for Learning-based Steering (NeurIPS 2021 · similarity 0.84)
- S-NeRF: Neural Radiance Fields for Street Views (ICLR 2023 · similarity 0.84)
- Learn TAROT with MENTOR: A Meta-Learned Self-Supervised Approach for Trajectory Prediction (ICCV 2023 · similarity 0.83)
- READ: Large-Scale Neural Scene Rendering for Autonomous Driving (AAAI 2023 · similarity 0.83)
- DualAD: Disentangling the Dynamic and Static World for End-to-End Driving (CVPR 2024 · similarity 0.83)
- ViiNeuS: Volumetric Initialization for Implicit Neural Surface Reconstruction of Urban Scenes with Limited Image Overlap (CVPR 2025 · similarity 0.82)
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.
