organized files and code

Author: Ubuntu

RESOURCES.md

@@ -0,0 +1,68 @@
+# %%
+import glob
+# read in fmri data
+import os
+import matplotlib.pyplot as plt
+import numpy as np
+import torch
+import pandas as pd
+import matplotlib.pyplot as plt
+
+# Read in video features
+VID_DIR = "/home/ubuntu/hcp_data/stimuli_jpg/"
+video1 = "7T_MOVIE1_CC1_v2_224x224_72.npy"
+video2 = "7T_MOVIE2_HO1_v2_224x224_72.npy"
+video3 = "7T_MOVIE3_CC2_v2_224x224_72.npy"
+video4 = "7T_MOVIE4_HO2_v2_224x224_72.npy"
+
+vid1_feat = np.load(VID_DIR + video1, allow_pickle=True)
+vid2_feat = np.load(VID_DIR + video2, allow_pickle=True)
+vid3_feat = np.load(VID_DIR + video3, allow_pickle=True)
+vid4_feat = np.load(VID_DIR + video4, allow_pickle=True)
+
+# %%
+
+# Choose the `slowfast_r50` model
+model = torch.hub.load('facebookresearch/pytorchvideo', 'slowfast_r50', pretrained=True)
+proj_layer = model.blocks[6].proj
+
+
+# %%
+
+df_labels = pd.read_csv("kinetics_400_labels.csv")
+
+labels = df_labels.name.values
+print(labels.shape)
+
+
+# %%
+
+videos = [video1,video2,video3,video4]
+video_raw_outputs = []
+video_classifications = []
+threshold = 0.80
+with torch.no_grad():
+    for video in videos:
+        vid_feat = np.load(VID_DIR + video, allow_pickle = True)
+        vid_feat_torch = torch.Tensor(vid_feat)
+        vid_feat_last = proj_layer(vid_feat_torch)
+        vid_classification = torch.sigmoid(vid_feat_last)
+        vid_classification[vid_classification >= threshold] = 1
+        vid_classification[vid_classification < threshold] = 0
+        video_raw_outputs.append(vid_feat_last)
+        video_classifications.append(vid_classification)
+        np.save((VID_DIR + video).replace("_72","_72_last_layer"), vid_feat_last.detach().numpy(), allow_pickle=True)
+        np.save((VID_DIR + video).replace("_72","_72_sigmoid"), vid_classification.detach().numpy(), allow_pickle=True)
+
+# %%
+
+
+
+frame = 1050
+labels_for_frame = labels[video_classifications[0][frame]==1]
+path_frames = "/home/ubuntu/hcp_data/stimuli_jpg/7T_MOVIE1_CC1_v2_224x224"
+path_frame = path_frames+ "/frame"+ str(frame+1).zfill(4) + ".jpg"
+img = plt.imread(path_frame)
+plt.imshow(img)
+plt.title(str(labels_for_frame))
+# %%