如何训练自己数的据集

IDEA-Research / GroundingDINO

[ECCV 2024] Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"

https://arxiv.org/abs/2303.05499

Apache License 2.0

6.6k stars 675 forks source link

Open Zhuohan007 opened 1 year ago

Zhuohan007 commented 1 year ago

作者您好，如何使用你们预训练好的GroundingDINO模型训练自己的数据集？（在自己的数据集数量很少的情况下）

SlongLiu commented 1 year ago

GroundingDINO 在fewshot情况下有很好的性能，您可以直接finetune已有模型。

脚本我们还在清理，可能在稍后放出。

Zhuohan007 commented 1 year ago

好的非常感谢！

xiexie123 commented 1 year ago

请问思路是固定其他，finetune Cross-Modality Decoder 吗

kinredon commented 1 year ago

@SlongLiu 请问什么时候会 release finetune 的脚本呢

DietDietDiet commented 1 year ago

@SlongLiu 同等待！加油清理

heyongxin233 commented 1 year ago

May I ask how long it will take to release the finetune script?

EddieEduardo commented 1 year ago

同求，感谢 in advance！！！

HaoLiuHust commented 1 year ago

when will you release the finetune code?

yinjun622 commented 1 year ago

期待

wojiaohumaocheng commented 1 year ago

什么时候出训练、finetune脚本谢谢

smile-care commented 1 year ago

等一手

imaginistLi commented 1 year ago

插眼