Clarifications (ongoing)

[mixmixmix]

This is just my little list of things to clarify about the setup (ongoing)

• - manualOveride.ipynb only allows for adding missing boxes, not for deleting them
• - ./tracking/yolo_detector.py from deep_sort.detection import Detection - how should I import deep sort? Are we actually using it there? @ctorney?

[ctorney]

you can delete by double-clicking in the box

[mixmixmix]

The deep_sort import is unnecessary. (I think :^)

[aakanksharathore]

When I run train.py for 2 training images and 1 epoch, it runs for a while and then stops with error message:

tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[32,864,864,32] and type float on /job:localhost/replica:0/task:0/device:CPU:0 by allocator cpu
[[{{node leaky_0/LeakyRelu/mul}} = Mul[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](leaky_104/LeakyRelu/alpha, bnorm_0/cond/Merge)]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

Does it have something to do with my CPU allocation?

[mixmixmix]

Can you share the images you work with on google drive or dropbox?
I have a few images from Colin and I don't have any problems "toy training" on my CPU (for 1 epoch)

[aakanksharathore]

Sure, I am attaching images and annotation in email if that's fine?

…

On Thu, Jan 24, 2019 at 12:59 PM M. Kunde ***@***.***> wrote: Can you share the images you work with on google drive or dropbox? I have a few images from Colin and I don't have any problems "toy training" on my CPU (for 1 epoch) — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ANIAIga1Qqv3TZUxywxtGpCUz-jcuoAdks5vGa5FgaJpZM4aPKh4> .

-- Regards, Aakanksha सर्वे भवन्तु सुखिनः सर्वे सन्तु निरामयाः । सर्वे भद्राणि पश्यन्तु मा कश्चिद्दुःखभाग्भवेत् । ॐ शान्तिः शान्तिः शान्तिः ॥

[mixmixmix]

What weights are you using?

[aakanksharathore]

yolo-v3-coco.h5

[mixmixmix]

Can run fine tuning using horses-yolo.h5? The training seem to work fine for me so it might have something to do with tensorflow setup.

[aakanksharathore]

I had set fine tuning as 0, was training only final layers. It may be something to do with my CPU and tensorflow. So, we are good to train it on the server then?

[mixmixmix]

If you want to do just final layers, the FINE_TUNING should be 1, methinks.

I think we can train on the server! The annotations look fine to me, we just need to correct the path once we upload the data (- filename: /home/aakanksha/Documents/uavTracker-master/train/yoloTrainData/frame209_6.png).

[mixmixmix]

I'm almost ready with the branch with annotations and configs in yml, which should make it a bit easier to train (since there is no need to change sourcecode each time, I hope.

[aakanksharathore]

Getting error while running "runTracker.py"

tracking/yolo_tracker.py", line 32, in
from filterpy.kalman import KalmanFilter
ImportError: No module named 'filterpy'

In uavTracker repository I couldn't find any file named filterpy

[aakanksharathore]

Hi Miks,

what should be the clip name in videos.yml file?

filename: "videos/04MarchMor2_02_03.avi"
transform: "tracks/04MarchMor2_02_03_MAT.npy"
periods:

• {start: 14400, stop: 23400, clipname: "04MarchMor2_02_03_clip1"}

Can I give any random name? Where will we use clipname?

[mixmixmix]

Clipname will be appended to the resulting filename. So to have output file 04MarchMor2_02_03_clip1_TR.avi you just need clipname clip1.

Sorry for the late reply!

[aakanksharathore]

Thank you. *Namastey!* Akanksha Rathore PhD scholar CES, Indian Institute of Science Bangalore 560012 website: https://sites.google.com/view/akanksha-rathore twitter: akanksha_ak5 <https://twitter.com/akanksha_ak5>

…

------------------------------------------------------------------------------------------------------------- सर्वे भवन्तु सुखिनः सर्वे सन्तु निरामयाः । सर्वे भद्राणि पश्यन्तु मा कश्चिद्दुःखभाग्भवेत् । ॐ शान्तिः शान्तिः शान्तिः ॥

On Sun, Jun 9, 2019 at 4:03 PM M. Kunde ***@***.***> wrote: Clipname will be appended to the resulting filename. So to have output file 04MarchMor2_02_03_clip1_TR.avi you just need clipname clip1. Sorry for the late reply! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=ADJAAIXWWLBMF6JFTZAKJR3PZTL7ZA5CNFSM4GR4VB4KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXIHUGA#issuecomment-500202008>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADJAAIXVIDAOCVTFC7MKVKLPZTL7ZANCNFSM4GR4VB4A> .

[mixmixmix]

Just referencing here issue which deals with two-phase fine tuning. #7