123 points by john_doe 4 months ago flag hide 12 comments
johnny5alive 4 months ago next
This is really cool! I've been waiting for a breakthrough in image recognition for a while now.
deeplearner123 4 months ago next
I tried it on my own image dataset and I'm getting really impressive results, especially with the transfer learning.
machinewhisperer 4 months ago next
Do you think this could be used for medical image analysis? I work on mammography images and this could be a game changer.
deeplearner123 4 months ago next
I think it could certainly be used for medical applications. I haven't tried it personally, but I'd be interested to hear your results.
sarahcodes 4 months ago next
Sure, I'll post results here when I get a chance to try it on mammography images.
deeplearner123 4 months ago next
Can't wait to see your results with medical images. I think this architecture's simplicity will make it more accessible to non-experts.
profgraham 4 months ago prev next
I've skimmed the paper and it looks promising. Has anyone tried it with their own datasets yet?
pytorchqueen 4 months ago next
Yes, I have a medium-sized image dataset for object detection and the results are impressive. I'll post a link to my experiments.
sarahcodes 4 months ago next
Here's the link to my GitHub repo with code and results: https://github.com/sarahcodes/neuralnet-image-recognition
pytorchqueen 4 months ago next
Thanks for sharing. I started implementing this architecture and it's not as difficult as I thought. Impressive work.
machinewhisperer 4 months ago next
I like the modularity of this approach. I've been looking for a simpler way to incorporate recent advancements in convolutional neural networks.
sarahcodes 4 months ago next
I'd be interested to see how well this architecture scales to bigger datasets. I'll try it out on ImageNet and post the results here.