Research map / Detection and segmentation / CNN features
Mobile Video Object Detection With Temporally-Aware Feature Maps
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 temporally-aware Bottleneck-LSTMs to MobileNet-SSD | MobileNet-SSD α=1 · ImageNet VID 2015 minival | mAP: 50.3 -> 54.4 | Helped |
| Add temporally-aware Bottleneck-LSTMs to MobileNet-SSD | MobileNet-SSD α=0.5 · ImageNet VID 2015 minival | mAP: 40.0 -> 43.8 | Helped |
| Add temporally-aware Bottleneck-LSTMs to MobileNet-SSD | MobileNet-SSD α=1 · ImageNet VID 2015 minival | parameters (M): 4.20 -> 3.24 | Helped |
| Add temporally-aware Bottleneck-LSTMs to MobileNet-SSD | MobileNet-SSD α=1 · ImageNet VID 2015 minival | MAC (B): 1.20 -> 1.13 | Helped |
| Place one LSTM after Conv13 rather than omit recurrence (ablation) | MobileNet-SSD α=1 · ImageNet VID 2015 minival | mAP: 50.3 -> 53.5 | Helped |
| Place one LSTM after Conv3 rather than omit recurrence (ablation) | MobileNet-SSD α=1 · ImageNet VID 2015 minival | mAP: 50.3 -> 49.1 | Hurt |
| Replace the Bottleneck-LSTM with weighted averaging (ablation) | MobileNet-SSD α=0.5 · ImageNet VID 2015 minival | mAP: 42.8 -> 37.6 | Hurt |
| Replace the Bottleneck-LSTM with a standard LSTM (ablation) | MobileNet-SSD α=1 · ImageNet VID 2015 minival | MAC (M), recurrent layer: 34 -> 840 | Hurt |
| Replace the Bottleneck-LSTM with a GRU (ablation) | MobileNet-SSD α=1 · ImageNet VID 2015 minival | mAP: 53.5 -> 54.0 | Helped |
| Add a Bottleneck-LSTM after Feature Map 1 in addition to Conv13 (ablation) | MobileNet-SSD α=1 · ImageNet VID 2015 minival | mAP: 53.5 -> 53.7 | Helped |
| Stack two LSTMs after the same Conv13 feature map (ablation) | MobileNet-SSD · ImageNet VID 2015 minival | mAP: 42.8 -> 42.9; 53.5 -> 53.3 | No clear effect |
| Increase the LSTM output width beyond α_lstm=0.25α (ablation) | Bottleneck-LSTM after Conv13 · ImageNet VID 2015 minival | mAP | Hurt |
Closest papers by meaning
- MobileDets: Searching for Object Detection Architectures for Mobile Accelerators (CVPR 2021 · similarity 0.88)
- Recurrent Vision Transformers for Object Detection With Event Cameras (CVPR 2023 · similarity 0.88)
- Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (CVPR 2017 · similarity 0.88)
- Context R-CNN: Long Term Temporal Context for Per-Camera Object Detection (CVPR 2020 · similarity 0.87)
- Rethinking Scale-Aware Temporal Encoding for Event-based Object Detection (NeurIPS 2025 · similarity 0.87)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.86)
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.