The Code for M2CAI19 Paper: Hard Frame Detection and Online Mapping for Surgical Phase Recognition.

Pytorch version >= 0.4 python version >= 3.0 GPU Memory >= 9000mb

We downsample the videos at 5fps. We recommand you download the dataset with this link: https://mega.nz/#!rEN2FCwB!VEChVilyX2yGpge_pjfjNuu8qH6qeFX6q5b8_v_ydVw https://mega.nz/#!zqZkGaBT!P5hhMe_BuyM3RQPRu0jdJR-x3EEmcB0cAe0BOxw5-_U

The folder 'DataCleaning' contains the code for data cleansing step. This step is time-costing, so, we suggest to use the data cleansing results that are contained in the link we provided.

The folder 'OHFM' contains the code for OHFM. Download the pretrained model and data, run the command:

 python main.py --dataset=cholec80-workflow-5 --savepath=test_result

utils_2.py: Code for drawing segment bars.