WIP: Feature/generate result video

.gitignore

@@ -1,4 +1,8 @@
 data/
 data2/
 __pycache__
-model_checkpoints/
+model_checkpoints/
+output/
+*.mp4
+*.png
+*.txt

data/extract_frames.py

@@ -10,6 +10,10 @@
 import datetime
 import argparse
 
+import numpy as np
+from PIL import Image
+import skvideo.io
+
 
 def extract_frames(video_path):
     frames = []

result.txt

@@ -0,0 +1,37 @@
+9 9 Mating_501.mp4
+205 217 Mating_502.mp4
+205 217 Mating_503.mp4
+392 483 Mating_504.mp4
+205 217 Mating_505.mp4
+243 308 Mating_506.mp4
+0 49 Sniffing_506.mp4
+15 56 Sniffing_507.mp4
+0 19 Sniffing_508.mp4
+0 6 Sniffing_509.mp4
+0 459 Sniffing_511.mp4
+6 36 Sniffing_512.mp4
+0 28 Sniffing_513.mp4
+28 62 Sniffing_514.mp4
+12 61 Sniffing_515.mp4
+16 67 Sniffing_516.mp4
+20 30 Sniffing_517.mp4
+5 48 Sniffing_518.mp4
+48 292 Sniffing_519.mp4
+2 89 Sniffing_520.mp4
+0 24 Sniffing_521.mp4
+18 91 Sniffing_522.mp4
+1 107 Sniffing_523.mp4
+0 94 Sniffing_524.mp4
+12 140 Sniffing_525.mp4
+1 78 Sniffing_526.mp4
+0 32 Sniffing_527.mp4
+11 122 Sniffing_528.mp4
+27 63 Sniffing_529.mp4
+30 45 Sniffing_530.mp4
+20 26 Sniffing_531.mp4
+22 82 Sniffing_532.mp4
+19 27 Sniffing_533.mp4
+0 115 Sniffing_534.mp4
+4 122 Sniffing_535.mp4
+18 58 Sniffing_536.mp4
+0 30 Sniffing_537.mp4

run_testOnVideo.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+# Define the input file and the checkpoint model path
+input_file="data/ucfTrainTestlist/testlist01.txt"
+checkpoint_model="model_checkpoints/ConvLSTM_195.pth"
+video_path_prefix="data/UCF-101/"
+
+# Check if the file exists
+if [ -f "$input_file" ]; then
+  # Read and process each video path in the input file
+  while IFS= read -r video_path; do
+    # Remove the '\r' character from the video path
+    video_path=$(echo "$video_path" | tr -d '\r')
+    # Add the video path prefix
+    full_video_path="${video_path_prefix}${video_path}"
+    # Execute the Python script with the video path and checkpoint model
+    python3 test_on_video.py --video_path "$full_video_path" --checkpoint_model "$checkpoint_model"
+  done < "$input_file"
+else
+  echo "Error: File '$input_file' does not exist."
+fi

static_graph.py

@@ -0,0 +1,49 @@
+import matplotlib.pyplot as plt
+
+# Read the data from the text file
+with open("result.txt", "r") as file:
+    data = [line.strip().split() for line in file.readlines()]
+
+# # Separate the data into durations, video names, and their indices
+# durations = [int(item[0]) for item in data]
+# video_names = [item[2] for item in data]
+# video_indices = range(len(video_names))
+
+# # Create a bar graph using matplotlib
+# plt.bar(video_indices, durations)
+# plt.xticks(video_indices, video_names, rotation=45, ha="right")
+# plt.xlabel("Video Names")
+# plt.ylabel("Duration")
+# plt.title("Duration of Videos")
+# plt.tight_layout()
+
+# # Save the graph to a file and display it
+# plt.savefig("output_graph.png")
+# plt.show()
+
+
+
+import matplotlib.pyplot as plt
+
+# Data
+files = [item[2] for item in data]
+actual = [int(item[0]) for item in data]
+expected = [int(item[1]) for item in data]
+
+# Calculate accuracy
+accuracy = [a / e * 100 for a, e in zip(actual, expected)]
+
+# Plot the bar graph
+fig, ax = plt.subplots()
+ax.bar(files, accuracy)
+
+# Set labels and title
+ax.set_ylabel('Accuracy (%)')
+ax.set_xlabel('Files')
+ax.set_title('Accuracy for Mating Files')
+
+# Rotate x-axis labels for better readability
+plt.xticks(rotation=45)
+
+# Show the graph
+plt.show()

test_on_video.py

@@ -8,6 +8,10 @@
 from torchvision.utils import make_grid
 from PIL import Image, ImageDraw
 import skvideo.io
+import numpy as np
+import cv2
+
+
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
@@ -38,13 +42,25 @@
     labels = sorted(list(set(os.listdir(opt.dataset_path))))
 
     # Define model and load model checkpoint
-    #model = ConvLSTM(input_shape=input_shape, num_classes=len(labels), latent_dim=opt.latent_dim)
-    model = ConvLSTM(num_classes=len(labels), latent_dim=opt.latent_dim)
+    ### model = ConvLSTM(input_shape=input_shape, num_classes=len(labels), latent_dim=opt.latent_dim)
+    model = ConvLSTM(
+        num_classes=len(labels),
+        latent_dim=opt.latent_dim,
+        lstm_layers=1,
+        hidden_dim=1024,
+        bidirectional=True,
+        attention=True,
+    )
     model.to(device)
-    #model.load_state_dict(torch.load(opt.checkpoint_model))
-    model.load_state_dict(torch.load(opt.checkpoint_model), strict=False)
+    model.load_state_dict(torch.load(opt.checkpoint_model))
     model.eval()
 
+    ### labels statistics
+    num_true_labels = 0
+    input_file = opt.video_path.split('/')
+    true_labels = input_file[2]
+    #print(type(opt.video_path), opt.video_path)
+
     # Extract predictions
     output_frames = []
     for frame in tqdm.tqdm(extract_frames(opt.video_path), desc="Processing frames"):
@@ -62,8 +78,48 @@
 
         output_frames += [frame]
 
-    # Create video from frames
-    writer = skvideo.io.FFmpegWriter("output.gif")
-    for frame in tqdm.tqdm(output_frames, desc="Writing to video"):
-        writer.writeFrame(np.array(frame))
-    writer.close()
+        ### Count correct predictions and save the result in text file
+        if true_labels == predicted_label:
+            num_true_labels += 1
+    # Save the result to a text file
+    output_file = "result.txt"
+    with open(output_file, "a") as file:
+        result = str(num_true_labels) + " " + str(len(output_frames)) + " " + str(input_file[3])
+        file.write(result + "\n")
+
+    # Print a confirmation message
+    print(f"result of '{input_file[3]}' has been saved to {output_file}.")
+    ### print(num_true_labels, len(output_frames))
+    # # Create video from frames
+    # writer = skvideo.io.FFmpegWriter("output.gif")
+    # for frame in tqdm.tqdm(output_frames, desc="Writing to video"):
+    #     writer.writeFrame(np.array(frame))
+    # writer.close()
+
+    # Create output folder if it does not exist
+    output_folder = "output"
+    if not os.path.exists(output_folder):
+        os.makedirs(output_folder)
+
+
+# Save output_frames as images
+    for i, frame in enumerate(output_frames):
+        frame_np = np.array(frame)
+        image = Image.fromarray(frame_np)
+        image.save(os.path.join(output_folder, f"frame_{i}.jpg"))
+
+    # # Set up VideoWriter object
+    # video_width, video_height = 640, 480
+    # fourcc = cv2.VideoWriter_fourcc(*"mp4v")
+    # video_writer = cv2.VideoWriter("output.mp4", fourcc, 25.0, (video_width, video_height))
+
+    # for i, frame in enumerate(output_frames):
+    #     # Convert PIL Image to numpy array
+    #     frame = np.array(frame)
+    #     # Resize frame to video size
+    #     frame = cv2.resize(frame, (video_width, video_height))
+    #     # Write frame to video file
+    #     video_writer.write(frame)
+
+    # # Release VideoWriter object
+    # video_writer.release()