changed visualizations

utils.py

@@ -1,10 +1,12 @@
 import torch
+import subprocess
 import numpy as np
 import supervision as sv
 import cv2
 import os
 from glob import glob
 from tqdm import tqdm
+import math
 
 
 def plot_predictions(
@@ -25,8 +27,13 @@ def plot_predictions(
         in zip(labels, scores)
     ]
 
-    bbox_annotator = sv.BoxAnnotator(color_lookup=sv.ColorLookup.INDEX)
-    label_annotator = sv.LabelAnnotator(color_lookup=sv.ColorLookup.INDEX)
+    height, width, _ = image_source.shape
+    thickness = math.ceil(width/200)
+    text_scale = width/1500
+    text_thickness = math.ceil(text_scale*1.5)
+
+    bbox_annotator = sv.BoxAnnotator(color_lookup=sv.ColorLookup.INDEX, thickness=thickness)
+    label_annotator = sv.LabelAnnotator(color_lookup=sv.ColorLookup.INDEX, text_scale=text_scale, text_thickness=text_thickness)
     annotated_frame = cv2.cvtColor(image_source, cv2.COLOR_RGB2BGR)
     annotated_frame = bbox_annotator.annotate(scene=annotated_frame, detections=boxes)
     annotated_frame = label_annotator.annotate(scene=annotated_frame, detections=boxes, labels=labels)
@@ -41,7 +48,8 @@ def mp4_to_png(input_path: str, save_path: str, scale_factor: float) -> str:
     # get frames per second
     fps = int(cv2.VideoCapture(input_path).get(cv2.CAP_PROP_FPS))
     # run subprocess to convert mp4 to pngs 
-    os.system(f"ffmpeg -i {input_path} -vf 'scale=iw*{scale_factor}:ih*{scale_factor}, fps={fps}' {save_path}/frame%08d.png")
+    os.system(f"ffmpeg -i {input_path} -vf 'fps={fps},scale=iw*{scale_factor}:ih*{scale_factor}' {save_path}/frame%08d.png")
+    # subprocess.run(["ffmpeg", "-i", input_path, "-vf", f"scale=iw*{scale_factor}:ih*{scale_factor}, fps={fps}", f"{save_path}/frame%08d.png"])
     return fps
 
 def vid_stitcher(frames_dir: str, output_path: str, fps: int = 30) -> str: