NNNNerd / Triple-I-Net-TINet

Official code for "Illumination-guided RGBT Object Detection with Inter- and Intra-modality Fusion"
Apache License 2.0
16 stars 2 forks source link

关于KAIST数据集 #10

Open xxwbe opened 2 months ago

xxwbe commented 2 months ago

作者,您好。我在FLIR_aligned数据集上跑出了结果,对于KAIST数据集,没有正确运行。 File "/root/autodl-tmp/Triple-I-Net-TINet/mmdet/datasets/samplers/group_sampler.py", line 36, in __iter__ indices = np.concatenate(indices) File "<__array_function__ internals>", line 5, in concatenate 可以上传一份模型的配置文件,以及JSON格式的标注吗,我想参考一下,谢谢!

NNNNerd commented 2 months ago

你好!配置文件在根目录下faster_rcnn_r50_fpn_tinet_kaist.py;JSON格式的标注已上传,见README。另外,MR指标的测试是通过上传检测结果文件,利用KAIST官方测试服务器完成的。

xxwbe commented 2 months ago

我们准备在论文中引用您的文章作为对比,拟采用的格式如下: Near | Medium | Far | None | Partial | Heavy | Day | Night | All

但是您的论文中KAIST数据集的实验结果只有(Day | Night | All)三个数据,冒昧请问下您可否将(Near | Medium | Far | None | Partial | Heavy )的实验数据也分享一下,可以发到邮箱(boenwang@mail.ustc.edu.cn),谢谢!