I am not able to load the .pkl files provided. I can load the .json tho. What I get from the .json file is an int number indicating the model_type. Anyone knows how to associate the int number of .pkl file?
", "messages": [ { "id": "658332", "postDate": "10/25/2019 22:33:27", "content": "I am not able to load the .pkl files provided. I can load the .json tho. What I get from the .json file is an int number indicating the model_type. Anyone knows how to associate the int number of .pkl file?
", "rawMarkdown": "I am not able to load the .pkl files provided. I can load the .json tho. What I get from the .json file is an int number indicating the model_type. Anyone knows how to associate the int number of .pkl file?", "votes": null }, { "id": "658614", "postDate": "10/26/2019 08:58:15", "content": "@yunxiangcd Did you mean .json files in car_models_json.zip? These files can be used to get not only *car_type* but faces and vertices. You can examine my kernel HERE and look at the 13th cell to see how I extract these properties from .json files.
", "rawMarkdown": "yunxiangcd Did you mean .json files in car_models_json.zip? These files can be used to get not only *car_type* but *faces* and *vertices*. You can examine my kernel [HERE](https://www.kaggle.com/phunghieu/a-quick-simple-eda/) and look at the 13th cell to see how I extract these properties from .json files.", "votes": null }, { "id": "658896", "postDate": "10/26/2019 17:52:00", "content": "Got it. Thank you !
", "rawMarkdown": "Got it. Thank you !", "votes": null }, { "id": "658916", "postDate": "10/26/2019 18:45:23", "content": "@phunghieu I can see you how extract the model and car type (\"SUV\" etc.) but do you know how to associate models with the integer model types given in the PredictionString column of train.csv?
\n\nSorry if I'm missing something obvious 😄
", "rawMarkdown": "phunghieu I can see you how extract the model and car type (\"SUV\" etc.) but do you know how to associate models with the integer model types given in the PredictionString column of train.csv?\n\nSorry if I'm missing something obvious 😄", "votes": null }, { "id": "658927", "postDate": "10/26/2019 19:04:13", "content": "You are not the only one looking for that information. I asked organizers here some days ago, but no answer so far.
", "rawMarkdown": "You are not the only one looking for that information. I asked organizers [here](https://www.kaggle.com/c/pku-autonomous-driving/discussion/113888#656069) some days ago, but no answer so far.", "votes": null }, { "id": "658942", "postDate": "10/26/2019 19:28:37", "content": "It looks like they may have lifted their 3D models from the ApolloScape dataset. If so, then this file contains a mapping of model names to IDs. At the very least the model names and types in that file line up with this competition's JSON files.
", "rawMarkdown": "It looks like they may have lifted their 3D models from the ApolloScape dataset. If so, then [this file](https://github.com/ApolloScapeAuto/dataset-api/blob/master/car_instance/car_models.py) contains a mapping of model names to IDs. At the very least the model names and types in that file line up with this competition's JSON files.", "votes": null }, { "id": "658954", "postDate": "10/26/2019 20:13:47", "content": "I just checked a couple of examples and that mapping doesn't seem to be correct, though I guess those might have been mis-labellings. Also, perhaps the data was labelled with the 3d model that is the closest match to the shape of the car and not the exact make?\nEdit: ignore this, looks like I was wrong :)
", "rawMarkdown": "I just checked a couple of examples and that mapping doesn't seem to be correct, though I guess those might have been mis-labellings. Also, perhaps the data was labelled with the 3d model that is the closest match to the shape of the car and not the exact make?\nEdit: ignore this, looks like I was wrong :)", "votes": null }, { "id": "659008", "postDate": "10/26/2019 23:02:36", "content": "Still no.. that is actually the question I was asking..Any idea?
", "rawMarkdown": "Still no.. that is actually the question I was asking..Any idea?", "votes": null }, { "id": "659035", "postDate": "10/27/2019 00:58:55", "content": "@ollieperree @yunxiangcd I don't know how to map these values, either 😂
", "rawMarkdown": "ollieperree @yunxiangcd I don't know how to map these values, either 😂", "votes": null }, { "id": "659056", "postDate": "10/27/2019 02:15:40", "content": "@ebouteillon @jsday96 @ollieperree @phunghieu Thanks to James. https://raw.githubusercontent.com/ApolloScapeAuto/dataset-api/master/car_instance/car_models.py will give the correct mapping.
", "rawMarkdown": "ebouteillon @jsday96 @ollieperree @phunghieu Thanks to James. https://raw.githubusercontent.com/ApolloScapeAuto/dataset-api/master/car_instance/car_models.py will give the correct mapping.", "votes": null }, { "id": "659062", "postDate": "10/27/2019 02:30:13", "content": "Nice!
", "rawMarkdown": "Nice!", "votes": null }, { "id": "681997", "postDate": "11/26/2019 19:20:27", "content": "Anyone got an answer as how to open these pkl files. What's in them anyways?
", "rawMarkdown": "Anyone got an answer as how to open these pkl files. What's in them anyways?", "votes": null }, { "id": "682026", "postDate": "11/26/2019 20:14:48", "content": "The pkl files are broken/impossible to open in kernals, use the JSON files
", "rawMarkdown": "The pkl files are broken/impossible to open in kernals, use the JSON files", "votes": null } ], "comments": [ { "id": 658614, "author_name": "phunghieu", "author_url": "", "post_date": "10/26/2019 08:58:15", "content": "@yunxiangcd Did you mean .json files in car_models_json.zip? These files can be used to get not only *car_type* but faces and vertices. You can examine my kernel HERE and look at the 13th cell to see how I extract these properties from .json files.
", "votes": null, "replies": [ { "id": 658896, "author_name": "yunxiangcd", "author_url": "", "post_date": "10/26/2019 17:52:00", "content": "Got it. Thank you !
", "votes": null, "replies": [] }, { "id": 658916, "author_name": "ollieperree", "author_url": "", "post_date": "10/26/2019 18:45:23", "content": "@phunghieu I can see you how extract the model and car type (\"SUV\" etc.) but do you know how to associate models with the integer model types given in the PredictionString column of train.csv?
\n\nSorry if I'm missing something obvious 😄
", "votes": null, "replies": [] }, { "id": 659008, "author_name": "yunxiangcd", "author_url": "", "post_date": "10/26/2019 23:02:36", "content": "Still no.. that is actually the question I was asking..Any idea?
", "votes": null, "replies": [] }, { "id": 659035, "author_name": "phunghieu", "author_url": "", "post_date": "10/27/2019 00:58:55", "content": "@ollieperree @yunxiangcd I don't know how to map these values, either 😂
", "votes": null, "replies": [] } ] }, { "id": 658927, "author_name": "ebouteillon", "author_url": "", "post_date": "10/26/2019 19:04:13", "content": "You are not the only one looking for that information. I asked organizers here some days ago, but no answer so far.
", "votes": null, "replies": [ { "id": 658942, "author_name": "jsday96", "author_url": "", "post_date": "10/26/2019 19:28:37", "content": "It looks like they may have lifted their 3D models from the ApolloScape dataset. If so, then this file contains a mapping of model names to IDs. At the very least the model names and types in that file line up with this competition's JSON files.
", "votes": null, "replies": [] }, { "id": 658954, "author_name": "ollieperree", "author_url": "", "post_date": "10/26/2019 20:13:47", "content": "I just checked a couple of examples and that mapping doesn't seem to be correct, though I guess those might have been mis-labellings. Also, perhaps the data was labelled with the 3d model that is the closest match to the shape of the car and not the exact make?\nEdit: ignore this, looks like I was wrong :)
", "votes": null, "replies": [] } ] }, { "id": 659056, "author_name": "yunxiangcd", "author_url": "", "post_date": "10/27/2019 02:15:40", "content": "@ebouteillon @jsday96 @ollieperree @phunghieu Thanks to James. https://raw.githubusercontent.com/ApolloScapeAuto/dataset-api/master/car_instance/car_models.py will give the correct mapping.
", "votes": null, "replies": [ { "id": 659062, "author_name": "phunghieu", "author_url": "", "post_date": "10/27/2019 02:30:13", "content": "Nice!
", "votes": null, "replies": [] } ] }, { "id": 681997, "author_name": "dmitriylit", "author_url": "", "post_date": "11/26/2019 19:20:27", "content": "Anyone got an answer as how to open these pkl files. What's in them anyways?
", "votes": null, "replies": [ { "id": 682026, "author_name": "greatgamedota", "author_url": "", "post_date": "11/26/2019 20:14:48", "content": "The pkl files are broken/impossible to open in kernals, use the JSON files
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "658332": "I am not able to load the .pkl files provided. I can load the .json tho. What I get from the .json file is an int number indicating the model_type. Anyone knows how to associate the int number of .pkl file?", "658614": "yunxiangcd Did you mean .json files in car_models_json.zip? These files can be used to get not only *car_type* but *faces* and *vertices*. You can examine my kernel [HERE](https://www.kaggle.com/phunghieu/a-quick-simple-eda/) and look at the 13th cell to see how I extract these properties from .json files.", "658896": "Got it. Thank you !", "658916": "phunghieu I can see you how extract the model and car type (\"SUV\" etc.) but do you know how to associate models with the integer model types given in the PredictionString column of train.csv?\n\nSorry if I'm missing something obvious 😄", "658927": "You are not the only one looking for that information. I asked organizers [here](https://www.kaggle.com/c/pku-autonomous-driving/discussion/113888#656069) some days ago, but no answer so far.", "658942": "It looks like they may have lifted their 3D models from the ApolloScape dataset. If so, then [this file](https://github.com/ApolloScapeAuto/dataset-api/blob/master/car_instance/car_models.py) contains a mapping of model names to IDs. At the very least the model names and types in that file line up with this competition's JSON files.", "658954": "I just checked a couple of examples and that mapping doesn't seem to be correct, though I guess those might have been mis-labellings. Also, perhaps the data was labelled with the 3d model that is the closest match to the shape of the car and not the exact make?\nEdit: ignore this, looks like I was wrong :)", "659008": "Still no.. that is actually the question I was asking..Any idea?", "659035": "ollieperree @yunxiangcd I don't know how to map these values, either 😂", "659056": "ebouteillon @jsday96 @ollieperree @phunghieu Thanks to James. https://raw.githubusercontent.com/ApolloScapeAuto/dataset-api/master/car_instance/car_models.py will give the correct mapping.", "659062": "Nice!", "681997": "Anyone got an answer as how to open these pkl files. What's in them anyways?", "682026": "The pkl files are broken/impossible to open in kernals, use the JSON files" }, "source": "meta" }