AI Paper Atlas

Research map / Video / Video segmentation and tracking

Video Task Decathlon: Unifying Image and Video Tasks in Autonomous Driving

Thomas E. Huang, Yifan Liu, Luc Van Gool, Fisher Yu

ICCV 2023 · Video › Video segmentation and tracking · approach: Physics-inspired motion prediction · 14 citations (OpenAlex)

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 changedWhereMeasuredEffect
naive joint multi-task training without CPFResNet-50 multi-task baseline · VTD validationVTDA: 218.2 -> 216.9Hurt
add CPF training protocol to multi-task baselineResNet-50 multi-task baseline · VTD validationVTDA: 216.9 -> 221.5Improved
add CPF training protocol to multi-task baselineSwin-T multi-task baseline · VTD validationVTDA: 217.0 -> 223.7Improved
add feature interaction blocks (VTDNet) to multi-task baselineResNet-50 · VTD validationVTDA: 221.5 -> 225.3Improved
add feature interaction blocks (VTDNet) to multi-task baselineSwin-T · VTD validationVTDA: 223.7 -> 226.9Improved
add Intra-IB (ablation)VTDNet ResNet-50 · VTD validationVTDA: 221.5 -> 223.9Improved
add Cross-IB on top of Intra-IB (ablation)VTDNet ResNet-50 · VTD validationVTDA: 223.9 -> 225.3Improved
add curriculum training (ablation)VTDNet ResNet-50 · VTD validationVTDA: 216.7 -> 217.8Improved
add pseudo-labels for pose and semantic segmentation (ablation)VTDNet ResNet-50 · VTD validationsemantic segmentation IoU: 42.0 -> 63.2Improved
add fine-tuning (ablation)VTDNet ResNet-50 · VTD validationVTDA: 222.4 -> 225.3Improved
double loss weights of one task group (ablation)VTDNet ResNet-50 · VTD validationVTDA: 225.3 -> 224.7No clear effect
scale backbone from ResNet-50 to ConvNeXt-BVTDNet · VTD validationVTDA: 225.3 -> 230.7Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy