Research map / 3D vision / Depth and stereo
Learning Depth from Single Monocular Images Using Deep Convolutional Neural Fields
Fayao Liu, Chunhua Shen, Guosheng Lin, Ian D. Reid
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 |
|---|---|---|---|
| jointly learn CRF pairwise parameters with the CNN (add CRF loss layer) (ablation) | DCNF · NYU v2 | rel error: 0.295 -> 0.257 | Improved |
| fine-tune the whole network after pre-training (ablation) | DCNF · NYU v2 | rel error: 0.257 -> 0.230 | Improved |
| fine-tune the whole network after pre-training (ablation) | DCNF · Make3D C1 | rel error: 0.331 -> 0.314 | Improved |
| use deeper CNN unary instead of SVR on CNN features (ablation) | unary only model · NYU v2 | rel error: 0.313 -> 0.295 | Improved |
| add hand-tuned smoothness term to SVR (ablation) | SVR · NYU v2 | rel error: 0.313 -> 0.290 | Improved |
| add hand-tuned smoothness term to unary only model (ablation) | Unary only · NYU v2 | rel error: 0.295 -> 0.287 | Improved |
| replace fully connected layer with fully convolutional net and superpixel pooling | DCNF · NYU v2 | rel error: 0.230 -> 0.237 | No clear effect |
| replace fully connected layer with fully convolutional net and superpixel pooling | DCNF · Make3D C1 | rel error: 0.314 -> 0.312 | No clear effect |
| use superpixel pooling in fully convolutional net | DCNF · NYU v2 | training speed: ~10 times faster | Improved |
| swap AlexNet part for VGG-16 | DCNF-FCSP · NYU v2 | rel error: 0.237 -> 0.213 | Improved |
| swap AlexNet part for VGG-16 | DCNF-FCSP · Make3D | rms error: C1 9.10 -> 7.36; C2 13.24 -> 14.09 | Mixed |
| add cross-bilateral filter post-processing | DCNF · NYU v2 | depth error | No clear effect |
Closest papers by meaning
- Monocular Depth Estimation Using Multi-Scale Continuous CRFs as Sequential Deep Networks (TPAMI 2018 · similarity 0.93)
- Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation (CVPR 2018 · similarity 0.93)
- Multi-Scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation (CVPR 2017 · similarity 0.92)
- Visualization of Convolutional Neural Networks for Monocular Depth Estimation (ICCV 2019 · similarity 0.92)
- Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue (ECCV 2016 · similarity 0.92)
- Unified Depth Prediction and Intrinsic Image Decomposition from a Single Image via Joint Convolutional Neural Fields (ECCV 2016 · similarity 0.92)
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.
