Hello everyone, I'm new to CNN models, and I have one very basic doubt, how to select or choose which pre-trained model to use for any given image classification task. It is just something to learn out of experience, or do we need to try roughly all pre-trained models, or there exists some rule of thumb or intuition thing. Thanks in advance
", "messages": [ { "id": 989944, "postDate": "2020-08-29T08:28:19.177Z", "content": "I would suggest you to use try all models you can because no model is best for all problems. This will also help to learn more as you are new to cnn models and the code and models you create can be used in future.
", "rawMarkdown": "I would suggest you to use try all models you can because no model is best for all problems. This will also help to learn more as you are new to cnn models and the code and models you create can be used in future.", "votes": 1, "replies": [ { "id": 992395, "postDate": "2020-08-31T06:12:26.917Z", "content": "Hi, thanks for advice. One doubt that whether we need black box knowledge or moderate knowledge or in-depth knowledge of pre-trained models, because I found their architecture very less intuitive, and anyways we can use them with keras without knowing too much about them ;-)
", "rawMarkdown": "Hi, thanks for advice. One doubt that whether we need black box knowledge or moderate knowledge or in-depth knowledge of pre-trained models, because I found their architecture very less intuitive, and anyways we can use them with keras without knowing too much about them ;-)" } ] }, { "id": 1005474, "postDate": "2020-09-10T14:17:09.680Z", "content": "Here is a list of easy to use pretrained model and their accuracy on ImageNet
", "rawMarkdown": "[Here ](https://keras.io/api/applications/)is a list of easy to use pretrained model and their accuracy on ImageNet" }, { "id": 989920, "postDate": "2020-08-29T08:07:20.510Z", "content": "Hello everyone, I'm new to CNN models, and I have one very basic doubt, how to select or choose which pre-trained model to use for any given image classification task. It is just something to learn out of experience, or do we need to try roughly all pre-trained models, or there exists some rule of thumb or intuition thing. Thanks in advance
", "rawMarkdown": "Hello everyone, I'm new to CNN models, and I have one very basic doubt, how to select or choose which pre-trained model to use for any given image classification task. It is just something to learn out of experience, or do we need to try roughly all pre-trained models, or there exists some rule of thumb or intuition thing. Thanks in advance" } ], "comments": [ { "id": 989944, "author_name": "Parth Chhabra", "author_url": "", "post_date": "2020-08-29T08:28:19.177000", "content": "I would suggest you to use try all models you can because no model is best for all problems. This will also help to learn more as you are new to cnn models and the code and models you create can be used in future.
", "votes": 1, "replies": [ { "id": 992395, "author_name": "tau__tsm1", "author_url": "", "post_date": "2020-08-31T06:12:26.917000", "content": "Hi, thanks for advice. One doubt that whether we need black box knowledge or moderate knowledge or in-depth knowledge of pre-trained models, because I found their architecture very less intuitive, and anyways we can use them with keras without knowing too much about them ;-)
", "votes": 0, "replies": [] } ] }, { "id": 1005474, "author_name": "Sven Prüß", "author_url": "", "post_date": "2020-09-10T14:17:09.680000", "content": "Here is a list of easy to use pretrained model and their accuracy on ImageNet
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "989944": "I would suggest you to use try all models you can because no model is best for all problems. This will also help to learn more as you are new to cnn models and the code and models you create can be used in future.", "1005474": "[Here ](https://keras.io/api/applications/)is a list of easy to use pretrained model and their accuracy on ImageNet", "989920": "Hello everyone, I'm new to CNN models, and I have one very basic doubt, how to select or choose which pre-trained model to use for any given image classification task. It is just something to learn out of experience, or do we need to try roughly all pre-trained models, or there exists some rule of thumb or intuition thing. Thanks in advance" } }