import os
from glob import glob
import pandas as pd
from sklearn.preprocessing import LabelEncoder
from .base import BaseIsolatedDataset
from ..data_readers import load_frames_from_video
[docs]class DeviSignDataset(BaseIsolatedDataset):
"""
Chinese Isolated Sign language dataset from the paper:
`The devisign large vocabulary of chinese sign language database and baseline evaluations`
"""
lang_code = "csl"
[docs] def read_glosses(self):
self.glosses = []
df = pd.read_csv(self.class_mappings_file_path, delimiter="\t", encoding="utf-8")
for i in range(len(df)):
self.glosses.append(df["Meaning (Chinese)"][i].strip())
[docs] def read_original_dataset(self):
"""
Check the file "DEVISIGN Technical Report.pdf" inside `Documents\` folder
for dataset format (page 12) and splits (page 15)
TODO: The train set size is 16k, and test set size is 8k (for 2k classes).
Should we use 4k from test set as valset, and only the other 4k for benchmarking?
"""
if "rgb" in self.modality:
common_filename = "color.avi"
elif "pose" in self.modality:
common_filename = "pose.pkl"
else:
raise NotImplementedError
if self.split_file:
df = pd.read_csv(self.split_file)
for i in range(len(df)):
video_path = df["video_path"][i]
video_file = os.path.join(self.root_dir, video_path, common_filename)
if not os.path.isfile(video_file):
raise FileNotFoundError(video_file)
gloss_id = int(video_file.replace("\\", "/").split("/")[-2].split("_")[1])
instance_entry = video_file, gloss_id
self.data.append(instance_entry)
else:
video_files_path = os.path.join(self.root_dir, "**", common_filename)
video_files = glob(video_files_path, recursive=True)
if not video_files:
exit(f"No videos files found for: {video_files_path}")
for video_file in video_files:
naming_parts = video_file.replace("\\", "/").split("/")[-2].split("_")
gloss_id = int(naming_parts[1])
signer_id = int(naming_parts[0].replace("P", ""))
if (signer_id <= 4 and "train" in self.splits) or (
signer_id > 4 and ("test" in self.splits or "val" in self.splits)
):
instance_entry = video_file, gloss_id
self.data.append(instance_entry)
[docs] def read_video_data(self, index):
video_name, label = self.data[index]
video_path = os.path.join(self.root_dir, video_name)
imgs = load_frames_from_video(video_path)
return imgs, label, video_name