Instructions to use Zeteng/CameraJev-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Zeteng/CameraJev-9B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("video-classification", model="Zeteng/CameraJev-9B")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Zeteng/CameraJev-9B", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download README.md from Zeteng/CameraJev-9B: direct link, hf CLI and curl.
- Browser
- Download file 3.03 kB
-
https://huggingface.co/Zeteng/CameraJev-9B/resolve/main/README.md
- Command line
-
hf download hf://Zeteng/CameraJev-9B/README.md
-
curl -L -H "Authorization: Bearer $HF_TOKEN" -o README.md https://huggingface.co/Zeteng/CameraJev-9B/resolve/main/README.md
license: cc-by-nc-sa-4.0
base_model: Qwen/Qwen3.5-9B
pipeline_tag: video-classification
library_name: transformers
language:
- en
- zh
tags:
- camera-motion
- video-understanding
- cinematography
- video-classification
- lora
extra_gated_heading: 申請取用 · Request access
extra_gated_prompt: >-
研究用 checkpoint,僅供**非商業研究**(CC-BY-NC-SA-4.0),逐個人工審批。
本倉庫**不含任何視頻或幀**。方法細節、評測結果與使用說明隨倉庫文件一併提供,審核通過後可見。
Research checkpoint for **non-commercial research only** (CC-BY-NC-SA-4.0),
approved
manually. **No video or frames are included.** Method details, evaluation
results and
usage instructions ship with the repository files and are visible once
approved.
extra_gated_fields:
姓名 Full name: text
所屬機構 Affiliation: text
用途說明 Intended use: text
我僅用於非商業研究 Non-commercial research only: checkbox
extra_gated_button_content: 申請取用 Request access
CameraJev-9B — camera-motion recognition
CameraJev-9B recognises how the camera moves in a video — pan, tilt, truck, pedestal, dolly, zoom, roll, arc, tracking, static.
It is built for scale: the model is designed to run over large video corpora rather than a handful of clips, and it holds up on real film footage, which is where camera-motion recognition is hardest.
Evaluated on the standard public camera-motion benchmarks. Quantitative results, the method, and runnable inference code are in the repository and become available once your access request is approved — this work is not yet published, so the public card is kept brief.
What you get
A LoRA checkpoint over Qwen/Qwen3.5-9B, the model code needed to load it, and inference
scripts. No video, frames or training data are included.
Access
Click Request access on this page and complete the form. Requests are reviewed individually.
Licence and credit
- Licence: CC-BY-NC-SA-4.0 — non-commercial research only.
- Base model:
Qwen/Qwen3.5-9B(Apache-2.0), not redistributed here. - Model code under
videojev_code/derives from VideoJev (ZhuYuhan/VideoJev-4B, Apache-2.0); itsLICENSEandNOTICEare retained in that directory.
@misc{qwen3,
title = {Qwen3 Technical Report},
author = {{Qwen Team}},
year = {2025},
eprint = {2505.09388},
archivePrefix = {arXiv}
}
Citation
If you use this repository, please cite it. The upstream works credited above should be cited as well.
@misc{lin2026camerajev9b,
author = {Zeteng Lin and Jing Tang},
title = {CameraJev-9B: camera-motion recognition},
year = {2026},
howpublished = {\url{https://huggingface.co/Zeteng/CameraJev-9B}}
}