generate_result.py

import os
import shutil
from pathlib import Path
from base.utils import get_filename_from_a_folder_given_extension, get_all_files_recursively_by_ext
import numpy as np
import csv
import pandas as pd


valence_result_path = "/misc/scratch11/6th-ABAW_Competition_2024/6th-ABAW_Competition_23Feb2024/save/ABAW5_LFAN_RCMA_setup_new_fold1_valence_seed3407/predict/extra/valence"
arousal_result_path = "/misc/scratch11/6th-ABAW_Competition_2024/6th-ABAW_Competition_23Feb2024/save/ABAW5_CAN_orig_setup_can_new_fold1_valence_seed3407/predict/extra/valence"

test_set_list_path = r"L:\ABAW5\Valence_Arousal_Estimation_Challenge_test_set_release.txt"
test_set_list = pd.read_csv(test_set_list_path, header=None).values[:, 0]

sample_path = r"L:\ABAW5\sample.txt"
sample_df = pd.read_csv(sample_path)
raw_result_list = [os.path.join(raw_result_path, fpath) for fpath in  os.listdir(raw_result_path)]

for fold in range(6):
    fold_path = os.path.join(raw_result_path, "fold" + str(fold))
    valences, arousals, images = [], [], []
    for trial in test_set_list:

        arousal_txt, valence_txt = get_all_files_recursively_by_ext(fold_path, "txt", trial)
        assert "arousal" in arousal_txt and "valence" in valence_txt

        arousal = pd.read_csv(arousal_txt).values
        valence = pd.read_csv(valence_txt).values

        assert len(arousal) == len(valence)
        length = len(arousal)

        sample_length = len(sample_df[sample_df['image_location'].str.match(trial + "/")])

        diff = sample_length - length


        if diff > 0:

            length = sample_length
            arousal = np.concatenate((arousal, np.repeat(arousal[-1], diff)[:, np.newaxis]))
            valence = np.concatenate((valence, np.repeat(valence[-1], diff)[:, np.newaxis]))
        elif diff < 0:
            length = sample_length
            arousal = arousal[:sample_length, :]
            valence = valence[:sample_length, :]

        print("{} has diff = {} to {}".format(trial, str(diff), str(sample_length)))

        image = []
        for i in range(length):
            image.append(trial + "/" + str(i+1).zfill(5) + ".jpg")

        valences.extend(valence)
        arousals.extend(arousal)
        images.extend(image)


    result_path = os.path.join(final_result_path, "fold" + str(fold) + ".txt")
    result = np.c_[images, valences, arousals]

    result_df = pd.DataFrame(result, columns=["image_location","valence","arousal"])
    result_df.to_csv(result_path,sep=",", index=None)
    print("fold {} done!".format(str(fold)))