I want to try ensemble SRNet and Efficient.\nBut SRNet's output is different format Efficient.\nHow to ensemble SRNet and Efficient?\nPlease help me.
", "messages": [ { "id": "940074", "postDate": "07/22/2020 17:00:29", "content": "I want to try ensemble SRNet and Efficient.\nBut SRNet's output is different format Efficient.\nHow to ensemble SRNet and Efficient?\nPlease help me.
", "rawMarkdown": "I want to try ensemble SRNet and Efficient.\nBut SRNet's output is different format Efficient.\nHow to ensemble SRNet and Efficient?\nPlease help me.", "votes": null }, { "id": "940126", "postDate": "07/22/2020 17:26:52", "content": "Since AUC is a metric that only depends on the order I think any number of methods would work.
\n\nI am using the rank order method described in this kernel by ragnar.
", "rawMarkdown": "Since AUC is a metric that only depends on the order I think any number of methods would work.\n\nI am using the [rank order method described in this kernel](https://www.kaggle.com/ragnar123/rank-then-blend) by [ragnar](https://www.kaggle.com/ragnar123).", "votes": null }, { "id": "942240", "postDate": "07/23/2020 16:48:08", "content": "A simple answer is You cannot use GAN and image classifiers in same context as both have different ways of using the data and getting output from them SRNet generates image like representations while Efficient nets generate a different subspace representation of images .\nAnother thing is if you're thing of generating more images from SRNet using the Training images , you are thinking different but SRNet is not particularly meant for this task, it is only useful to remove or replace a text, not hairs or anything which I think you want to remove from certain images.\nOnly way of using GAN's together would be to generate more similar dimensional images from train set(similar to Augmentations but a little advanced) and then train model on this combined data.
", "rawMarkdown": "A simple answer is You cannot use GAN and image classifiers in same context as both have different ways of using the data and getting output from them SRNet generates image like representations while Efficient nets generate a different subspace representation of images .\nAnother thing is if you're thing of generating more images from SRNet using the Training images , you are thinking different but SRNet is not particularly meant for this task, it is only useful to remove or replace a text, not hairs or anything which I think you want to remove from certain images.\nOnly way of using GAN's together would be to generate more similar dimensional images from train set(similar to Augmentations but a little advanced) and then train model on this combined data.", "votes": null } ], "comments": [ { "id": 940126, "author_name": "pcjimmmy", "author_url": "", "post_date": "07/22/2020 17:26:52", "content": "Since AUC is a metric that only depends on the order I think any number of methods would work.
\n\nI am using the rank order method described in this kernel by ragnar.
", "votes": null, "replies": [] }, { "id": 942240, "author_name": "bharatsahu", "author_url": "", "post_date": "07/23/2020 16:48:08", "content": "A simple answer is You cannot use GAN and image classifiers in same context as both have different ways of using the data and getting output from them SRNet generates image like representations while Efficient nets generate a different subspace representation of images .\nAnother thing is if you're thing of generating more images from SRNet using the Training images , you are thinking different but SRNet is not particularly meant for this task, it is only useful to remove or replace a text, not hairs or anything which I think you want to remove from certain images.\nOnly way of using GAN's together would be to generate more similar dimensional images from train set(similar to Augmentations but a little advanced) and then train model on this combined data.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "940074": "I want to try ensemble SRNet and Efficient.\nBut SRNet's output is different format Efficient.\nHow to ensemble SRNet and Efficient?\nPlease help me.", "940126": "Since AUC is a metric that only depends on the order I think any number of methods would work.\n\nI am using the [rank order method described in this kernel](https://www.kaggle.com/ragnar123/rank-then-blend) by [ragnar](https://www.kaggle.com/ragnar123).", "942240": "A simple answer is You cannot use GAN and image classifiers in same context as both have different ways of using the data and getting output from them SRNet generates image like representations while Efficient nets generate a different subspace representation of images .\nAnother thing is if you're thing of generating more images from SRNet using the Training images , you are thinking different but SRNet is not particularly meant for this task, it is only useful to remove or replace a text, not hairs or anything which I think you want to remove from certain images.\nOnly way of using GAN's together would be to generate more similar dimensional images from train set(similar to Augmentations but a little advanced) and then train model on this combined data." }, "source": "meta" }