TensorFlow Tutorial 3: Object Detection Walk-through

!!! NOTE !!!

Due to some change Google made, it is necessary to do this fix for the moment:

This is due to some change Google has made, I haven't had a chance to research the specifics yet. I'll work out a more permanent fix for this eventually.

TensorFlow Tutorial 3: Object Detection Walk-through

GitHub repo:

[ionmedicaldesigns]

Fantastic tutorials. You approach the material with clear and concise methodologies. I've learned more from you than any other resource I've located to-date. You cover specific pipeline algorithms and that's exactly what I was looking for. I enjoy the history of TensorFlow and image/ object recognition as well as other considerations; however, when I just want to see how a set of test images perform with image or object recognition this is the resource I use, and recommend. Thanks for taking the time to produce these resources as I know it's a decent endeavor.

[gl]

Good day Chris, Thank you very much for the excellent explains for this difficult topic. Now it's accessible for me and with healthy study by myself it must be a road that lead to success. Thank you again and have a nice day!!!

[huynhthiensach]

Really exciting as waiting the trainning. For my PC, it takes 300 seconds for a step. Thank you so much!

[deephousefridays]

thank you for the tutorial, it was really great to watch it! Haven't seen so good step by step explaining in a long time! Totally appriciate it.

[StartupSignals]

THANK YOU CHRIS!!! Such an amazing use of learning time. More in depth TF tutorials would be love <3

[michaelmutekeri]

great video, only at the end when running the train script using tensorflow 2.0 it clashes
Traceback (most recent call last):
WARNING: The TensorFlow contrib module will not be included in TensorFlow 2.0.
But great video and planning, so far the best series I have done. Better than even paid courses I have been doing. I have been following the series from the first one.

[qzorn]

hi Chris, i was thinking about some of your interesting videos that i watched and learned a lot from as about opencv and raspberry pi stuff like color tracking... thanks please make more

[jacobusstrydom]

Thanks, what a damn good tutorial. Allot better that all the other once I have seen.

[Elzelgator]

dude I really liked your videos good work

[rio]

Very good Video, thanks for sharing, this helps me a lot. please consider to have a "Live Camera" demo for this.

[vivekranjan]

Really a Great tutorial..much appreciated..waiting for the .net implementation for the same..

[photorealm]

Thank you very much for sharing this information. You are one of the best I have ever seen . I followed your tutorial and every thing worked perfectly. I do have a question, Do the training images need the xml annotation file if all of the training images are cropped well? My thought is no but I can not find any proof of this any where.

[dadawong]

Great! Chris, do a speech recognition tutorial please!

[steveboyle]

Fantastic TF series. I'm having a discrepancy between the # of objects ID'd in jupyter (vs. the # in Your comments seem to indicate they are 2 ways to accomplish the same thing but I'm not getting the same results. Expected or unexpected? Thx

[baodo]

Awesome Chris. Can you show how to extract the bounding box coordinates for each detection ?

[prakashgadupoodi]

Hi Chris,
While I am running train.py file I am getting mscoco labelmap error.

I think you did not come across in video how to change labelmap.pbtxt file.

[guruprasaadh]

the tutorial worked fine...but when i tried it for my custom object detection i got lot of bounding boxes for 1 object do i do ....and the other labels didnt get recognized....and test images of other categories also showed the first label ....what do i do ?

[onghaingo]

Just wanna add some code has changed if you guys get errors in utils, utils => object_detection.utils.

[canozcivelek]

Hi! Thanks for the awesome tutorial series on object detection. I've got a small problem with it, hope you'll be able to find a solution. So I have 2 classes of objects to detect and I edited the labelmap, num_classes, and a few other files that were stated in the document accordingly. However, when I run test.py script, it only detects one of them and never shows the other. Where am I missing I was wondering...

[hongpinglo]

Thanks, it works like a charm, but i ran into a problem, when i test it on some very busy image, the detector only detect 20 object at most, is it possible to change some config to let the network detect more object(in my case around 45).