how to get ground-truth.txt and detection-result.txt

[linchunmian]

Hi, that's good job. I have confused that how to get ground-truth and detection result txt file? Does there provide script so that I could directly use to convert my ground truth label? I use KITTI vehicle data to evaluate yolov3 performance. The format of label records coordinate of all ground-truth bbox together, but I don't know how to separate them. And how can I convert test result to the format of detection-result.txt, just like the file in your repos? Could you please help me？

Thanks a lot in advance!

[WaqarAliMemon]

Same Question, I am using Tensorflow Object Detection API and it is giving me Output in the form of Images. How can i convert these images to detection.txt format. Any help would be appreciated.

[linchunmian]

Also, other question is how to get the ground truth file of test data and convert to txt format? Yow know, test data in most dataset don’t provide label information.I cannot to continute my research work now.发自我的iPhone------------------ Original ------------------From: WaqarAliMemon [email protected]Date: Wed,Oct 9,2019 1:21 PMTo: Cartucho/mAP [email protected]Cc: ChunmianLin [email protected], Author [email protected]Subject: Re: [SPAM] Re: [Cartucho/mAP] how to get ground-truth.txt anddetection-result.txt (#89)Same Question, I am using Tensorflow Object Detection API and it is giving me Output in the form of Images. How can i convert these images to detection.txt format. Any help would be appreciated.

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or mute the thread. [ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/Cartucho/mAP/issues/89?email_source=notifications\u0026email_token=AMWSBR7YW25S6RVANT5G5MLQNVS4NA5CNFSM4I2GOTJKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAWT5KQ#issuecomment-539836074", "url": "https://github.com/Cartucho/mAP/issues/89?email_source=notifications\u0026email_token=AMWSBR7YW25S6RVANT5G5MLQNVS4NA5CNFSM4I2GOTJKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAWT5KQ#issuecomment-539836074", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[WaqarAliMemon]

Do you have xml files? It can be converted to. Txt file using simple coding.

[linchunmian]

Test data usually cannot provide ground truth, does it?发自我的iPhone------------------ Original ------------------From: WaqarAliMemon [email protected]Date: Wed,Oct 9,2019 2:32 PMTo: Cartucho/mAP [email protected]Cc: ChunmianLin [email protected], Author [email protected]Subject: Re: [SPAM] Re: [Cartucho/mAP] how to get ground-truth.txt anddetection-result.txt (#89)Do you have xml files? It can be converted to. Txt file using simple coding.

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or mute the thread. [ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/Cartucho/mAP/issues/89?email_source=notifications\u0026email_token=AMWSBR2BLH254FLYQ5RGEM3QNV3GHA5CNFSM4I2GOTJKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAWZDQA#issuecomment-539857344", "url": "https://github.com/Cartucho/mAP/issues/89?email_source=notifications\u0026email_token=AMWSBR2BLH254FLYQ5RGEM3QNV3GHA5CNFSM4I2GOTJKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAWZDQA#issuecomment-539857344", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[6monkeyrs]

Hi, that's good job. I have confused that how to get ground-truth and detection result txt file? Does there provide script so that I could directly use to convert my ground truth label? I use KITTI vehicle data to evaluate yolov3 performance. The format of label records coordinate of all ground-truth bbox together, but I don't know how to separate them. And how can I convert test result to the format of detection-result.txt, just like the file in your repos? Could you please help me？

Thanks a lot in advance!

Hi guys, I will give it a try and hope that I don't confuse you all.

If you have not already seen Cartucho's OpenLabeling tool, please go here: https://github.com/Cartucho/OpenLabeling

Step1: The method I created detect-results text format files required for the mAP script, was to use OpenLabel script call main_auto.py script. Once I made this work and use my frozen inference model, I copied the script with a different name "main_auto_map.py" and added confidence list/numpy arrays to the functions where it creates the XML files.

Step 2: In this folder: https://github.com/Cartucho/mAP/tree/master/scripts/extra there is script to create XML to txt files for your ground truths. What I did is copy this script "convert_gt_xml.py" and name it "convert_dt_xml.py" because the scripts need to read the confidence level tag you add from step 1 script. This will create your required text files for the mAP.

I hope that this answers your question.