4 Region Of Interest (RoI Pooling) | Faster R-CNN | Object Detection Using Faster R-CNN

[@nandhininallamuthu4060]

This series of Faster R-CNN is amazing and well explained. Thanks for sharing the details!

[@CodeWithAarohi]

Welcome

[@soravsingla8782]

Amazing video ma’am. Please keep sharing

[@CodeWithAarohi]

Sure 😊

[@yugpn6358]

Yes It was helpful. Thanks for sharing !!

[@CodeWithAarohi]

Glad it was helpful!

[@justinamichael316]

Clear and superb explanation...

[@CodeWithAarohi]

Thank you 🙂

[@sridhurkeshsivaprakasam8591]

At first thanks for this video. Best explanation with implementation that i ever seen in youtube. Could you please tell how to train faster rcnn model?...Because i am more confused about that part

[@CodeWithAarohi]

Sure, I will do video on that part soon.

[@curious523]

what is the input_rois fed to the classifier layer? Where do we define it? is it the output of RPN? if yes, should I call rpn_layer or rpn_to_roi to get value for input_rois?

[@pifordtechnologiespvtltd5698]

Nicely explained

[@CodeWithAarohi]

Thank you so much 🙂

[@pifordtechnologiespvtltd5698]

Excellent

[@CodeWithAarohi]

Thanks

[@ratangles820]

Hi! Thanks for this video!! Is there any video after this one? That wraps everything to use it? You are amazing! Thanks!!

[@CodeWithAarohi]

welcome... You can check this video: kzbin.info/www/bejne/mnKXoYl9hpacqc0

[@camilamoro7025]

Hi Aarohi, Thank you for the video series it has been so helpful. I am a bit stuck with the testing part. Would you simply iteratively apply the functions in the rpn_layer and classifier files?

[@uni1357]

thanks for the great videos and explanations. Will you do any video on the testing phase of Faster RCNN? Thank you.

[@CodeWithAarohi]

Will try to do

[@shivaagrawal6115]

Hello Aarohi, thanks a lot for this wonderful video. One question, during the ROI pooling, why num_roi = 4 is selected ?

[@brunogreen25]

Thank you VERY MUCH for this very nice explanation of Faster RCNN's implementation. Could you, also, please specify which Tensorflow/Keras versions were you using?

[@CodeWithAarohi]

Glad my video helped you. I am using python 3.6.8, tensorflow 1.15.0, keras 2.2.5 and h5py version 2.10.0

[@jawadhaidar3931]

thank you for the video. I think there is a mistake in the RIOpooling function, the output should be the resized feature maps and not the resized image

[@CodeWithAarohi]

Thank you for bringing this to my attention. I appreciate your diligence and attention to detail. I will check that

[@pifordtechnologiespvtltd5698]

Excellent #BestAILearningChannel #CodeWithAarohi

[@mahesh1234m]

Good Explanation Madam. But Could you please add NMS also which is part of ROI pooling.

[@CodeWithAarohi]

Actually I have explained NMS in yolo algorithm, that's why not repeated here

[@curious523]

Hi Aarohi, I'm trying to replace the classifier of faster RCNN with KNN. After flattening the ROI pooling layer output, I need to convert it to numpy array and then pass it to KNN. The problem is I can not convert kerasTensor to NumPy array. I tried many ways. Does anyone have any ideas? thank you in advance

[@CodeWithAarohi]

Without trying I can't tell you what exactly the issue is?

[@parthapratimbarman7020]

Thank you so much, ma'am. Can you please show the training and testing part also with some loss function to get some training accuracy after running for some epochs?

[@mehnaztabassum1878]

Could you pls give the link to the RCNN test part? Thanks for the nice explanation.

[@CodeWithAarohi]

Hadn’t made video on that part yet

[@omarabubakr6524]

9:59 can I play with the number of dense layers or it is static in faster r-cnn

[@Ray-yh8lx]

the github link you provided , has 3 files . now how to use them for detection purpose , like the training testing etc part. please share that

[@bangarrajumuppidu8354]

clear cut explanation madam

[@CodeWithAarohi]

Thank you 🙂

[@teetanrobotics5363]

Amazing content Aarohi. Thank you. Could you please reorder the videos in the Deep Learning and the object detection playlists in the correct order?

[@CodeWithAarohi]

Sure, Will reorder the videos of my Playlist

[@TheAravindppadman]

why it is (nb_classes - 1) in out_regr instead of nb_classes?

[@CodeWithAarohi]

-1 because we don't want to work for background class

[@sarithapodali3782]

Is there a reason why the model would result in 0 positive anchors?

[@CodeWithAarohi]

I think you need to train model for more epochs to solve this problem.

[@sarithapodali3782]

@@CodeWithAarohi Thank you for responding. I think I have some positive anchors now. That was very helpful!

[@anjalinaidukarrotu1431]

File "", line 26 out_class = Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes)(out) ^ SyntaxError: can't assign to function call please tell me what is the error

[@KumarHemjeet]

Can we use polygonal annotations in faster RCNN?

[@ranjan561]

can we use this model for numerical data ?

[@CodeWithAarohi]

While Faster R-CNN was originally designed for object detection in images, it can be used for other applications as well, including numerical data. However, using Faster R-CNN for numerical data may not be the best choice in most cases, as Faster R-CNN is optimized for object detection in images and may not perform well on numerical data.

[@JackLee-oq2io]

could you share code de finetuning the detectron2_yolov7, please? Thanks

[@CodeWithAarohi]

1st notebook is github.com/AarohiSingla/Detectron2-Tutorial/blob/main/Detectron_fasterrcnn_pretrained_coco.ipynb and 2 notebook have only 2 commands to run pip install alfred-py and pip install nbnb

[@MovieMania-04]

Mam what we pass in base path ?

[@CodeWithAarohi]

Base path is the convolutional network which you use for feature extraction

[@dhanalakshmispillai9969]

I have send you rpy to your mail...

[@CodeWithAarohi]

checked and replied

[@dayanamary5103]

Tamil sollunga mam

[@dhanalakshmispillai9969]

Could you please check out the mail!

[@CodeWithAarohi]

Hi, I have checked my email id but not any mail

[@aresagathos2075]

This video might be very good, but i cant watch it, because of the voice :( i'm so sorry. Not even when i take a shot every time she says "Aroay".

[@CodeWithAarohi]

Sorry for the inconvenience.I will redo this video

[@aresagathos2075]

@@CodeWithAarohi Please don't redo the video, just some randoom guy complains about accent. A lot of people like your video. Right?

[@CodeWithAarohi]

@@aresagathos2075 ok

[@jesusgomez1280]

@@CodeWithAarohi No need to apologize or redo video. The video is not the problem. The problem is with so much unprofessional and racist people out there.

[@jesusgomez1280]

You should take classes on diversity and inclusion! 😡

[@HUEHUEUHEPony]

oh my god the accent

[@jesusgomez1280]

You should take classes on diversity and inclusion!😡

[@vemulasuman6995]

Excellent

[@CodeWithAarohi]

Thanks