Skip to page content
SAVRN
Search Plan a project

Independent publisher

chi-cheng Chen

SamChen567

Models in Library0
Datasets in Library1
Models on Hugging Face9
Followers—

Compare the published evidence

This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.

Published tasks: Visual question answering.

Reported licenses: other. Open a record to review its specific license and conditions.

Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.

Datasets

Dataset · Visual question answering

Video-XL-Pro-Union

chi-cheng Chen

依據以下兩篇論文整理的 SFT 訓練資料,兩者取聯集;兩者重疊的資料集一律採用 Video-XL-Pro 版本的標註。 標註只收錄兩個官方 repo(MINT-SJTU/Video-XL-Pro-Training、sy1998/VideoXLTraining)中有的檔案,且只保留標註實際引用到的媒體。論文或 repo 中沒有的資料集都已移除。Video-XL 的 pretraining 資料(LAION-2M 子集)因品質不佳不收錄。 訓練時把 image folder / video folder 設為 media/ 即可。所有標註路徑都已用 tools/verifypaths.py 驗證過,缺檔數為 0。 - activitynetqa15k.json 改指向 LLaVA178k/ActivityNet-QA/...(與 LLaVA-Video-178K 為同一批影片)。 - finevideo43k.json:移除原始檔中 35 筆 conversations 為空的樣本(43,751 → 43,716)。 - vcg107k.json:359 部影片實際為.mkv /.webm,路徑副檔名已改成實際檔名(影響 6,122 筆樣本)。 - FineVideo、ScienceQA 的原始檔名被作者改成 uuid:FineVideo 用 QA 內容完整比對對回原影片;ScienceQA 用題目、答案文字加上 uuid1 時間順序對回原圖(順序為 test → train → validation,對齊無歧義)。對應方式見 tools/finevideomap.py。…

Publicly accessible other 1M<n<10M