YACWC

SearchScratch/test_filter.py

@@ -0,0 +1,61 @@
+import pickle
+
+
+cache_files = ['/mnt/hdd_24tb_1/videos/ftp/leopards1/2025/09/12/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/mnt/hdd_24tb_1/videos/ftp/leopards1/2025/09/13/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/14/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/15/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/16/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/17/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/18/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl', '/srv/ftp_tcc/leopards1/2025/09/19/embedding_scores@0.0@926895f71538e3683e9af0956af94cf4.pkl']
+import time
+from datetime import timedelta, datetime
+start_time = time.time()
+all_c = list()
+
+start_time = 1757892175.042
+end_time =   1757894197.548
+
+def check_if_overlap(start_1, end_1, start_2, end_2):
+    ff = sorted([[start_1, end_1],[start_2, end_2]],key=lambda x: x[0])
+    return ff[0][1] > ff[1][0]
+
+def get_cache_data(start_time, end_time, cache_files):
+    targvals = [start_time, end_time]
+    for f in cache_files:
+        fold_start_time = datetime(*[int(x) for x in f.split('/')[-4:-1]]).timestamp()
+        fold_end_time = fold_start_time + 86400.0
+
+        has_overlap = check_if_overlap( start_time, end_time, fold_start_time, fold_end_time)
+
+        if not has_overlap:
+            continue
+
+        print(f'Loading {f}')
+
+        with open(f,'rb') as ff:
+            all_c.append(pickle.load(ff))
+    return all_c
+
+st = time.time()
+all_cach = get_cache_data(start_time, end_time, cache_files)
+
+vids = list()
+for c_c in all_cach:
+    vids.extend(c_c['videos'])
+
+data_filt = list()
+for v in vids:
+    if check_if_overlap( v['start_time'], v['end_time'], start_time, end_time):
+        data_filt.append(v)
+
+
+        
+time_vec = np.hstack([ np.asarray(f['embed_scores']['time'])+f['start_time'] for f in data_filt])
+score_vec = np.hstack([f['embed_scores']['score'] for f in data_filt])
+
+
+s_time, s_ind = np.unique(time_vec, return_index=True)
+s_score = score_vec[s_ind]
+
+
+out_array = np.asarray([s_time, s_score]).T.tolist()
+
+
+
+
+