Pre-trained Models: Urban vs Natural Images
\nI do not have much experience developing deep learning models in any environment other than with images from nature. For example I have created models for the Cassava Leaf Disease Competition and for the Plant Pathology Competition which is currently ongoing.
\nBoth of these competitions were focused on finding diseased leaves/plants and I found that the efficient-net and res-net models worked best in these competitions.
\n--
\nDoes anyone have any thoughts/advice for working with images in an urban environment? Do people often gravitate to certain pre-trained models in situations similar to these?
\n--
\nThanks to @usharengaraju for starting an awesome discussion here on relevant research papers regarding AI + Human Trafficking as well.
", "messages": [ { "id": 1271724, "postDate": "2021-04-12T20:56:43.513Z", "content": "Pre-trained Models: Urban vs Natural Images
\nI do not have much experience developing deep learning models in any environment other than with images from nature. For example I have created models for the Cassava Leaf Disease Competition and for the Plant Pathology Competition which is currently ongoing.
\nBoth of these competitions were focused on finding diseased leaves/plants and I found that the efficient-net and res-net models worked best in these competitions.
\n--
\nDoes anyone have any thoughts/advice for working with images in an urban environment? Do people often gravitate to certain pre-trained models in situations similar to these?
\n--
\nThanks to @usharengaraju for starting an awesome discussion here on relevant research papers regarding AI + Human Trafficking as well.
", "rawMarkdown": "**Pre-trained Models: Urban vs Natural Images** \n\nI do not have much experience developing deep learning models in any environment other than with images from nature. For example I have created models for the [Cassava Leaf Disease Competition](https://www.kaggle.com/c/cassava-leaf-disease-classification) and for the [Plant Pathology Competition](https://www.kaggle.com/c/plant-pathology-2021-fgvc8) which is currently ongoing. \n\nBoth of these competitions were focused on finding diseased leaves/plants and I found that the efficient-net and res-net models worked best in these competitions. \n\n--\n\nDoes anyone have any thoughts/advice for working with images in an urban environment? Do people often gravitate to certain pre-trained models in situations similar to these?\n\n--\n\nThanks to @usharengaraju for starting an awesome discussion [here](https://www.kaggle.com/c/hotel-id-2021-fgvc8/discussion/225145) on relevant research papers regarding AI + Human Trafficking as well. ", "votes": 1 } ], "comments": [], "raw_markdown_by_id": { "1271724": "**Pre-trained Models: Urban vs Natural Images** \n\nI do not have much experience developing deep learning models in any environment other than with images from nature. For example I have created models for the [Cassava Leaf Disease Competition](https://www.kaggle.com/c/cassava-leaf-disease-classification) and for the [Plant Pathology Competition](https://www.kaggle.com/c/plant-pathology-2021-fgvc8) which is currently ongoing. \n\nBoth of these competitions were focused on finding diseased leaves/plants and I found that the efficient-net and res-net models worked best in these competitions. \n\n--\n\nDoes anyone have any thoughts/advice for working with images in an urban environment? Do people often gravitate to certain pre-trained models in situations similar to these?\n\n--\n\nThanks to @usharengaraju for starting an awesome discussion [here](https://www.kaggle.com/c/hotel-id-2021-fgvc8/discussion/225145) on relevant research papers regarding AI + Human Trafficking as well. " } }