Which dataset the result is shown for?

Question

Which dataset the result is shown for?

Opened this issue 4 years ago · 9 comments

Hi,
Thank you for sharing the great work!

I have one question about the dataset for evaluation in your result table. The validation dataset here is the full set of validation (6852 images) or only a subset of 1000 selected images? (as I knew about KITTI dataset, there are two types of validation dataset like that)

Best,
Khang Truong.

Answer 1 · 2020-06-19T03:08:57.000Z

Hi,
Thank you for sharing the great work!

I have one question about the dataset for evaluation in your result table. The validation dataset here is the full set of validation (6852 images) or only a subset of 1000 selected images? (as I knew about KITTI dataset, there are two types of validation dataset like that)

Best,
Khang Truong.

Hi Khang Truong,

In the result table, we validate our model with 1000 selected images (the 'depth_selection' dataset in KITTI).

Ang

Answer 2 · 2020-06-19T04:09:29.000Z

Hi @anglixjtu ,

I want to evaluate your model on the full validation dataset. I need to write a script on my own for data loader, loading your pre-trained model and evaluation, right? Your source code only provides the script for the network architecture. Can you provide the full source code so that I can reproduce a reliable result?

Answer 3 · 2020-06-19T06:08:29.000Z

Hi @TruongKhang,

I will upload the full source code in these two days, and I will let you know. Or if you prefer, I can evaluate our model on the full validation dataset from my end and tell you the results.

Answer 4 · 2020-06-19T06:24:43.000Z

Thank you so much @anglixjtu . It would be very helpful for the research community when you provide the result on the full validation dataset. I'm looking forward to hear from you. :D

Answer 5 · 2020-06-19T06:39:28.000Z

Hi @TruongKhang,

I just update the results on the full validation dataset. Please check out README. Hope you find this information useful!

Answer 6 · 2020-06-19T06:45:14.000Z

Yes, thank you so much @anglixjtu .

Answer 7 · 2020-06-23T14:06:16.000Z

Hi @anglixjtu , when can you provide the full source code? I want to run your approach on the other datasets!

Answer 8 · 2020-06-24T03:22:19.000Z

Hi @TruongKhang，

I apologize that it's taken us a little longer than expected to prepare the code and made you wait for long. I just finished uploading the full source code. Try python3 run_eval.py -chkpt "/PATH/TO/YOUR/CHECKPOINT_FILE.pth.tar" to evaluate the model and python3 train.py to train the model.

Answer 9 · 2020-06-24T06:47:53.000Z

Thank you for sharing great work, @anglixjtu