# tinghuiz/SfMLearner

An unsupervised learning framework for depth and ego-motion estimation from monocular videos

Repository: https://github.com/tinghuiz/SfMLearner
Canonical: https://ross.abutalabs.com/products/sfmlearner
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: deep-learning, depth-prediction, visual-odometry, self-supervised-learning, unsupervised-learning
Last push: 2021-10-26T05:58:09+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": 3417, "days_push": 1772, "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 2017, forks 554 (observed 2026-08-28T04:06:06.132108+00:00)

## What it is
SfMLearner is a TensorFlow implementation of the CVPR 2017 paper 'Unsupervised Learning of Depth and Ego-Motion from Video'. It trains models to predict single-view depth and camera ego-motion from monocular video sequences without ground-truth labels.

## Use cases
- estimate depth from monocular video with a neural network
- train a self-supervised depth prediction model on KITTI
- predict camera ego-motion from image sequences
- run single-view depth prediction demo on pretrained model
- learn visual odometry without ground truth poses
- reproduce unsupervised SfM learning research results

## When to choose
- you need a research baseline for unsupervised depth and pose estimation
- you have monocular video from KITTI or Cityscapes and no labeled depth data
- you want to study or extend self-supervised structure-from-motion learning

## When to avoid
- you need production-ready depth estimation with modern architectures
- you require recent TensorFlow or GPU stack support (built for TF 1.0 / CUDA 8)
- you need real-time visual odometry for deployment rather than research

## Facets
- artifact type: library
- maturity: maintenance
- function: deep-learning, computer-vision, machine-learning
- domain: computer-vision, deep-learning, autonomous-vehicles, robotics
- platform: python
- tags: depth-estimation, ego-motion, visual-odometry, self-supervised-learning, monocular-video, tensorflow, research-code, kitti, cityscapes, linux, gpu

## Member repositories
- tinghuiz/SfMLearner (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:06.132108+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 2026-08-30T03:00:41.362909+00:00, confidence not recorded.
  - readme: https://github.com/tinghuiz/SfMLearner (fetched 2026-08-28T04:06:06.132108+00:00, sha 6a650c867575)
- Data as of 2026-08-30T08:39:29.467469+00:00.