# msracver/Deep-Feature-Flow

Deep Feature Flow for Video Recognition

Repository: https://github.com/msracver/Deep-Feature-Flow
Canonical: https://ross.abutalabs.com/products/deep-feature-flow
Language: Python
License: MIT
License Family: permissive
Last push: 2021-09-27T02:40:23+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3401, "days_push": 1801, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1315, forks 292 (observed 2026-08-28T04:04:20.581522+00:00)

## What it is
Official MXNet implementation of Deep Feature Flow (CVPR 2017), an end-to-end framework for video recognition such as object detection and semantic segmentation in videos. It speeds up recognition by running heavy networks on sparse key frames and propagating feature maps to other frames via a lightweight optical flow network.

## Use cases
- speed up video object detection with key-frame feature propagation
- semantic segmentation on video frames
- train video recognition models end-to-end with sparse annotations
- reproduce Deep Feature Flow research results
- implement optical-flow-based feature propagation in MXNet

## When to choose
- you need a fast, end-to-end trainable video recognition pipeline based on MXNet
- you want to reproduce or extend the CVPR 2017 DFF paper
- your video dataset has sparse frame-level annotations

## When to avoid
- you use PyTorch or TensorFlow rather than legacy MXNet
- you need actively maintained, production-ready code
- you need real-time inference on CPU-only hardware

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, video-processing
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: video-recognition, object-detection, semantic-segmentation, optical-flow, mxnet, cvpr-2017, research-code, video, linux, gpu

## Member repositories
- msracver/Deep-Feature-Flow (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:20.581522+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:48:42.406263+00:00, confidence not recorded.
  - readme: https://github.com/msracver/Deep-Feature-Flow (fetched 2026-08-28T04:04:20.581522+00:00, sha ecbe9be45e4b)
- Data as of 2026-08-30T08:39:29.467469+00:00.