Research map / Video / Video segmentation and tracking
Flow-Event Autoencoder: Event Stream Object Recognition Dataset Generation with Arbitrary High Temporal Resolution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a ViT encoder-decoder (flow-event-flow) that generates event streams from an image and optical flow at arbitrary temporal resolution, to augment DVS object recognition datasets.
Problem
Event-camera object recognition lacks large benchmark event-stream datasets; existing simulators (ESIM/vid2e, v2e, EventGAN) and display-based recording have refresh-rate, temporal resolution, noise, or tuning limitations.
Approach
Train a ViT encoder to render events per timestamp from an image and interpolated optical flow along a saccade trajectory, with a frozen pretrained event-to-flow decoder whose flow-reconstruction loss aligns generated events to realistic flow, enabling arbitrary temporal resolution generation.
Key results
- No experimental results are reported; the paper describes planned evaluation only (expectation that the method outperforms ESIM-vid2e and matches or beats N-ImageNet saccade data on N-Caltech101).
Limitations
- No experiments or quantitative results are presented; the work is a proposal in future tense.
- Training data plan is small (hundreds of printed images), and the paper gives no details on architecture size or training hyperparameters.
- Relies on a pretrained event-to-flow decoder whose quality bounds the generated events.
Experimental setup
Datasets: ImageNet N-ImageNet-1k N-Caltech101
Benchmarks: N-Caltech101
Metrics: object recognition accuracy after pretrain-transfer
Baselines: ESIM-vid2e N-ImageNet-1k (saccade recording) EventGAN v2e
Builds on: ESIM v2e EventGAN Orchard saccade-based N-MNIST recording
Closest papers by meaning
- Event Stream-based Visual Object Tracking: A High-Resolution Benchmark Dataset and A Novel Baseline (CVPR 2024 · similarity 0.85)
- Graph Neural Network Combining Event Stream and Periodic Aggregation for Low-Latency Event-based Vision (CVPR 2025 · similarity 0.84)
- Bilateral Event Mining and Complementary for Event Stream Super-Resolution (CVPR 2024 · similarity 0.84)
- Efficient Meshflow and Optical Flow Estimation from Event Cameras (CVPR 2024 · similarity 0.84)
- E2HQV: High-Quality Video Generation from Event Camera via Theory-Inspired Model-Aided Deep Learning (AAAI 2024 · similarity 0.84)
- PASS: Path-selective State Space Model for Event-based Recognition (NeurIPS 2025 · similarity 0.84)
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.