the standard of pkl

[pokameng]

Hello,I want to ask to you about the dateset's standard of pkl,for example the ucfl101...,because i want to train this code in charades.

[ArchiZX]

You can refer to this

MOC-Detector/src/datasets/dataset/dataset_annotation.py

Line 13 in 45c3fd0

class TubeDataset(object):

.

[pokameng]

我的意思是，我用UCF101的pkl，读出来的数据和我自己数据集的pkl格式完全不一样，我想问一下，就是您的pkl是怎么转换的。我是把您的pkl转成了json然后我把我的对应也转成了json，转换完，我再转回了pkl，但是跑程序的时候发现完全不一样

…

---Original--- From: "Zixu ***@***.***> Date: Mon, Apr 5, 2021 00:11 AM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [MCG-NJU/MOC-Detector] the standard of pkl (#29) You can refer to this https://github.com/MCG-NJU/MOC-Detector/blob/45c3fd0c5e4ca455e7e739f3ad80b2c91813df35/src/datasets/dataset/dataset_annotation.py#L13. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[pokameng]

您好，请问您能不能分享一下，生成pkl文件的代码？我生成的pkl格式和您的不太一样

…

---Original--- From: "Zixu ***@***.***> Date: Mon, Apr 5, 2021 00:11 AM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [MCG-NJU/MOC-Detector] the standard of pkl (#29) You can refer to this https://github.com/MCG-NJU/MOC-Detector/blob/45c3fd0c5e4ca455e7e739f3ad80b2c91813df35/src/datasets/dataset/dataset_annotation.py#L13. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[yixuanli98]

Sorry, we do not transform the GT pkl file and can not share the code. The GT pkl format is listed below:
Our GT pkl file is a dictionary with the following keys:

labels: list of labels

train_videos: a list with nsplits elements, each one containing the list of training videos

test_videos: idem for the test videos

nframes: dictionary that gives the number of frames for each video

resolution: dictionary that output a tuple (h,w) of the resolution for each video

gttubes: dictionary that contains the gt tubes for each video. Gt tubes are dictionary that associates from each index of label, a list of tubes. A tube is a numpy array with nframes rows and 5 columns: frame number, x1, y1, x2, y2.

You can transform your own dataset's GT file into this format.