CAMotion

CAMotion

A High-Quality Dataset for Camouflaged Motion Object Detection in the Wild

CAMotion is a dataset and benchmark for realistic camouflaged motion object detection in unconstrained, in-the-wild scenarios. It aims to support robust evaluation and reproducible research for motion camouflage understanding.

Paper GitHub Download
Under review • Last updated:
CAMotion dataset overview figure
Figure: Overview of CAMotion scenarios and annotations.

Overview

What is CAMotion?

CAMotion provides curated in-the-wild videos with camouflaged moving objects and pixel-level annotations. The benchmark focuses on motion camouflage challenges such as low contrast, background blending, complex motion, and occlusions.

Task

Given a video, predict a binary mask per frame for the camouflaged moving object(s). The dataset supports training and evaluation for video camouflage understanding and segmentation-style pipelines.

Highlights
  • Realistic in-the-wild motion camouflage scenarios
  • Frame-level pixel annotations for moving camouflaged objects
  • Standard splits and baseline results for reproducibility

Dataset Statistics

REPLACE
Videos
REPLACE
Frames
REPLACE
Categories

Splits

Split
Videos
Frames
Notes
Train
REPLACE
REPLACE
Standard training split
Val
REPLACE
REPLACE
Validation split
Test
REPLACE
REPLACE
Held-out evaluation split

Tip: Replace the values above with the official counts from your paper/dataset release.

Benchmark

Metrics

  • Mean IoU (mIoU)
  • F-measure / S-measure (if applicable)
  • MAE (if applicable)

Replace metrics list to match your paper’s evaluation protocol.

Baselines

We provide baseline implementations and evaluation scripts for reproducible comparison. See the GitHub repository for training instructions, checkpoints, and dataset loader.

Repository • Dataset Download

Results Table (Template)

Method
Backbone
mIoU
MAE
Link
Baseline-A
REPLACE
REPLACE
REPLACE
Baseline-B
REPLACE
REPLACE
REPLACE
Ours
REPLACE
REPLACE
REPLACE

You can keep this as a template or fill it with the official numbers.

Download

Dataset

Please use the links below to download the dataset and annotations. If you mirror the files, keep the directory structure unchanged for compatibility with the official loader.

Dataset (videos)
REPLACE size · ZIP / TAR
Annotations
REPLACE size · ZIP
Splits & Metadata
REPLACE size · TXT / JSON

By downloading CAMotion, you agree to use it for research purposes only and cite the paper.

Citation

Please cite CAMotion if you find it useful in your research:

@article{camotion,
  title   = {CAMotion: A High-Quality Dataset for Camouflaged Motion Object Detection in the Wild},
  author  = {Siyuan Yao and Hao Sun and Hai Long and Ruiqi Yu and Jiehong Li and Xiwei Jiang and Yanzhao Su and Wenqi Ren and Xiaochun Cao},
  journal = {Under review},
  year    = {2026}
}
Paper • GitHub