I'm looking for ways to train a OD model on Tensorflow. I tried to treat this problem as a segmentation problem and used segmentation models but that didn't work. And I would like to compile the model myself rather than using the TF OD API.
\nSo how can I make an Object detection model on TF?
", "messages": [ { "id": "1607177", "postDate": "12/05/2021 17:23:06", "content": "I'm looking for ways to train a OD model on Tensorflow. I tried to treat this problem as a segmentation problem and used segmentation models but that didn't work. And I would like to compile the model myself rather than using the TF OD API.
\nSo how can I make an Object detection model on TF?
", "rawMarkdown": "I'm looking for ways to train a OD model on Tensorflow. I tried to treat this problem as a segmentation problem and used segmentation models but that didn't work. And I would like to compile the model myself rather than using the TF OD API. \n\nSo how can I make an Object detection model on TF?", "votes": null }, { "id": "1612712", "postDate": "12/09/2021 08:44:15", "content": "I am coming from PyTorch and was also looking for a way to build object detection models. Since I am most interested in transfer learning, TensorFlow Hub seems to be a very nice platform to find pre-trained networks, that can be fine tuned with your own images. Another way to do this is to find implementations of popular OD models e.g. on Model Zoo. You can omit the weights and train them yourself if you like. Of course you can also come up with your own implementation. One way to do it is to turn a CNN image classifier into an object detector see here But there are many other ways to do it. I hope this helps a bit.
", "rawMarkdown": "I am coming from PyTorch and was also looking for a way to build object detection models. Since I am most interested in transfer learning, [TensorFlow Hub](https://tfhub.dev/s?module-type=image-object-detection) seems to be a very nice platform to find pre-trained networks, that can be fine tuned with your own images. Another way to do this is to find implementations of popular OD models e.g. on [Model Zoo](https://modelzoo.co/category/computer-vision). You can omit the weights and train them yourself if you like. Of course you can also come up with your own implementation. One way to do it is to turn a CNN image classifier into an object detector [see here](https://www.pyimagesearch.com/2020/06/22/turning-any-cnn-image-classifier-into-an-object-detector-with-keras-tensorflow-and-opencv/) But there are many other ways to do it. I hope this helps a bit.", "votes": null } ], "comments": [ { "id": 1612712, "author_name": "efrodl", "author_url": "", "post_date": "12/09/2021 08:44:15", "content": "I am coming from PyTorch and was also looking for a way to build object detection models. Since I am most interested in transfer learning, TensorFlow Hub seems to be a very nice platform to find pre-trained networks, that can be fine tuned with your own images. Another way to do this is to find implementations of popular OD models e.g. on Model Zoo. You can omit the weights and train them yourself if you like. Of course you can also come up with your own implementation. One way to do it is to turn a CNN image classifier into an object detector see here But there are many other ways to do it. I hope this helps a bit.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1607177": "I'm looking for ways to train a OD model on Tensorflow. I tried to treat this problem as a segmentation problem and used segmentation models but that didn't work. And I would like to compile the model myself rather than using the TF OD API. \n\nSo how can I make an Object detection model on TF?", "1612712": "I am coming from PyTorch and was also looking for a way to build object detection models. Since I am most interested in transfer learning, [TensorFlow Hub](https://tfhub.dev/s?module-type=image-object-detection) seems to be a very nice platform to find pre-trained networks, that can be fine tuned with your own images. Another way to do this is to find implementations of popular OD models e.g. on [Model Zoo](https://modelzoo.co/category/computer-vision). You can omit the weights and train them yourself if you like. Of course you can also come up with your own implementation. One way to do it is to turn a CNN image classifier into an object detector [see here](https://www.pyimagesearch.com/2020/06/22/turning-any-cnn-image-classifier-into-an-object-detector-with-keras-tensorflow-and-opencv/) But there are many other ways to do it. I hope this helps a bit." }, "source": "meta" }