285 - Object detection using Mask RCNN (with XML annotated data)

Code generated in the video can be downloaded from here:

This video helps you get started with your first Mask RCNN project that uses existing annotated data. Annotated data used in this video can be downloaded from:

XML data in PascalVOC format
XML annotation file for each image

Please note that this video assumes you have installed Mask RCNN on your system. For installation instructions, please watch the previous video (Video numbered 284).

[harastring]

Thanks for the video! Always exciting XD

[junaidmalik]

Can u tell me what version used for tensirflow and keras, I'm having an error " No Module name Keras.engine"

[lennard]

How would I add custom transformations?

Like this with detecto:

custom_transforms = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize(800),
transforms.ColorJitter(saturation=0.3),
transforms.ToTensor(),
utils.normalize_transform(),
])

dataset = core.Dataset('train_labels.csv', 'images/',
transform=custom_transforms)

[TruDeinoz]

21:40 I have the same issue. Weights were saved in a different location in a folder called logdir. How can I fix this?

[infosection]

First of all very thankful sir, your videos deserve more than millions of likes, sir, I'm working on segmentation, I have dataset images labeled with XML files, and my data set is about roads, I'm confused about using a U-NET model with this dataset. waiting for your kind response. Thanks

[jackharper]

I keep getting an error where it says

NotImplementedError: Exception encountered when calling Lambda.call().

We could not automatically infer the shape of the Lambda's output. Please specify the `output_shape` argument for this Lambda layer.

Do you know why I am getting this problem?

[PiyushASupe]

Would like to request MaskRCNN like series for mobilenetSSD, with Raspi deployment.

[lennard]

Is it possible NOT to detect something? Usually we only train with images where there is at least on object it should detect... however I noticed that the network detects objects with high confidence in completely different testing images although it is not supposed to.

[Programlama]

How can we see train/test results? I've found very useful to your videos but you are testing them with only single photo.

[estebanmeza]

can you use annotations made with labelimg?

[fredbaafi]

Thank you sir, your videos are really helpful, but my question is how do you assign the class when doing object detection on number plates, is it 2. That's 1 for car and 2 for number plate. Please i need help understanding it. Thank you

[sameerjasra]

Great explanation as usual. At your convenience can we have videos on Transformers. Thanks

[ayxanxasmmmdov]

Hi, when i train model, it returns nan value. How do i solve this problem?

[CRTagadiya]

Is Object detection series in on way? just like Image segmentation

[ashwiniyadav]

Sir can you suggest any mask for selecting particular pixel in an image

[MrBisht-ytou]

is your playlist is enough for ML and DL for beginner to intermediate level ?
what should I expect from this playlist ? Is the playlist still relevant today?

[DVonVal]

Hi, if i train 50 epoch first with coco weights, them can i replace coco weights with my last 50 epoch weights to train next 50 more epoch ? Or should i train straight 100 epoch? Or can i do to train with my last 50 epoch?

[mayaltaee]

Thanks sir, I hope you can make videos of Faster R-CNN for objects detection.

[masamatovic]

Can we use Ipad for picture labeling in APEER.. Thanks for the answer.. :)

[mujeebpa]

Can't find the code in GitHub Repo.