The Artificial Intelligence Forum                             data mining forum
This forum is to discuss artificial intelligence. You can ask anything about artificial intelligence : science-fiction, robots, consciousness, movies and so on! Also, you can discuss anything related to implementing artificial intelligence in machines such as programming, algorithms, etc. No registration is required!  
How to put a dynamic head on a Faster Rcnn
Posted by: Sam Jackson
Date: January 03, 2022 07:18PM

This is the repo https://github.com/microsoft/DynamicHead. I am trying to implement this with pytorch. Not via command line, but I am new and could use some instruction, if you please. Thank you.

Here is my code that used when creating my model:

self.backbone = Backbone()

self.backbone.out_channels = self.backbone.model.num_features

self.anchor_generator = AnchorGenerator(sizes=((32, 64, 128, 256, 512),),
aspect_ratios=((0.5, 1.0, 2.0),))

self.roi_pooler = torchvision.ops.MultiScaleRoIAlign(featmap_names=['0','1','2','3'],
output_size=14,
sampling_ratio=2)

self.model = FasterRCNN(self.backbone,
num_classes=2,
rpn_anchor_generator=self.anchor_generator,
box_roi_pool=self.roi_pooler)

Options: ReplyQuote


This forum is powered by Phorum and provided by P. Fournier-Viger (© 2012).
Terms of use.