How to reproduce the "average results over multiple runs"?

[FrankLeeCode]

To my understanding, the "Average results over multiple runs" are derived from utilizing 30 distinct sample seeds, as demonstrated in the data split. However, upon examining the re-organized data split, I noticed only one seed annotation provided (as indicated here). Consequently, I can only obtain results from a single run. How can I replicate the process to achieve the "Average results over multiple runs"?

[wangchen1801]

Thank you for your interest in our work! Currently, mmfewshot only support one fixed few shot dataset. To evaluate the samples generate from different seeds, we can replace the .txt in benchmark_[K]shot mannuly, and modify the img path, for example:

datasets/VOC2012/JPEGImages/2008_006761.jpg -> VOC2012/JPEGImages/2008_006761.jpg 

We can also register the different samples into FewShotVOCDataset in mmfewshot/mmfewshot/detection/datasets/voc.py, as the following code:

    voc_benchmark = {
        f'SPLIT{split}_{shot}SHOT': [
            dict(
                type='ann_file',
                ann_file=f'data/few_shot_ann/voc/benchmark_{shot}shot/'
                f'box_{shot}shot_{class_name}_train.txt',
                ann_classes=[class_name])
            for class_name in VOC_SPLIT[f'ALL_CLASSES_SPLIT{split}']
        ]
        for shot in [1, 2, 3, 5, 10] for split in [1, 2, 3]
    }

[FrankLeeCode]

Thank you for your prompt response! I truly appreciate it. Your explanation has resolved my query perfectly.