Skip to content

Commit 531aa7a

Browse files
bluemellophonebluemellophone
committed
Bug fixes for macOS
1 parent b20110f commit 531aa7a

1 file changed

Lines changed: 91 additions & 67 deletions

File tree

‎batbot/spectrogram/__init__.py‎

Lines changed: 91 additions & 67 deletions
Original file line numberDiff line numberDiff line change
@@ -137,7 +137,8 @@ def plot_histogram(
137137
hist_original = (hist_original / hist_original.max()) * hist.max()
138138

139139
hist_ = np.argmax(hist)
140-
hist_std = np.abs(image - hist_).mean()
140+
image_ = image.astype(np.int32)
141+
hist_std = np.abs(image_ - hist_).mean()
141142

142143
csum = np.cumsum(hist) / hist.sum()
143144
csum_ = np.where(csum >= csum_threshold)[0].min()
@@ -166,7 +167,7 @@ def plot_histogram(
166167
[0, y_max],
167168
color='grey',
168169
linestyle='--',
169-
label=f'Histogram Peak ({hist_:0.01f})',
170+
label=f'Histogram Peak ({hist_:0.01f} +/- {hist_std:0.01f})',
170171
)
171172
plt.plot(
172173
[csum_] * 2,
@@ -1307,6 +1308,7 @@ def compute_wrapper(
13071308
bands = bands[::-1]
13081309
y_step_freq = float(bands[0] - bands[1])
13091310
x_step_ms = float((1e3 * duration) / stft_db.shape[1])
1311+
bands = np.around(bands).astype(np.int32).tolist()
13101312

13111313
# # Save the spectrogram image to disk
13121314
# cv2.imwrite('debug.tif', stft_db, [cv2.IMWRITE_TIFF_COMPRESSION, 1])
@@ -1346,6 +1348,8 @@ def compute_wrapper(
13461348
}
13471349
metas = []
13481350
for index, (start, stop) in tqdm.tqdm(list(enumerate(ranges))):
1351+
if index == 1:
1352+
break
13491353
segment = stft_db[:, start:stop]
13501354

13511355
# Step 0.1 - Debugging setup and find peak amplitude (will return None if disabled)
@@ -1424,7 +1428,7 @@ def compute_wrapper(
14241428

14251429
# Step 11 - Collect chirp metadata
14261430
metadata = {
1427-
'curve.(khz,ms)': [
1431+
'curve.(hz,ms)': [
14281432
(
14291433
bands[y],
14301434
(start + x) * x_step_ms,
@@ -1438,38 +1442,40 @@ def compute_wrapper(
14381442
'fc.ms': (start + bands[fc[1]]) * x_step_ms,
14391443
'hi fc:knee.ms': (start + bands[knee[1]]) * x_step_ms,
14401444
'lo fc:heel.ms': (start + bands[heel[1]]) * x_step_ms,
1441-
'bandwidth.khz': bandwidth,
1442-
'hi f.khz': bands[top],
1443-
'lo f.khz': bands[bottom],
1444-
'peak f.khz': bands[peak[0]],
1445-
'fc.khz': bands[fc[0]],
1446-
'hi fc:knee.khz': bands[knee[0]],
1447-
'lo fc:heel.khz': bands[heel[0]],
1445+
'bandwidth.hz': bandwidth,
1446+
'hi f.hz': bands[top],
1447+
'lo f.hz': bands[bottom],
1448+
'peak f.hz': bands[peak[0]],
1449+
'fc.hz': bands[fc[0]],
1450+
'hi fc:knee.hz': bands[knee[0]],
1451+
'lo fc:heel.hz': bands[heel[0]],
14481452
'harmonic.flag': harmonic_flag,
14491453
'harmonic peak f.ms': (start + hamonic_peak[1]) * x_step_ms if harmonic_flag else None,
1450-
'harmonic peak f.khz': bands[hamonic_peak[0]] if harmonic_flag else None,
1454+
'harmonic peak f.hz': bands[hamonic_peak[0]] if harmonic_flag else None,
14511455
'echo.flag': echo_flag,
14521456
'echo peak f.ms': (start + echo_peak[1]) * x_step_ms if echo_flag else None,
1453-
'echo peak f.khz': bands[echo_peak[0]] if echo_flag else None,
1457+
'echo peak f.hz': bands[echo_peak[0]] if echo_flag else None,
14541458
}
14551459
metadata.update(slopes)
14561460

14571461
# Normalize values
14581462
for key, value in list(metadata.items()):
14591463
if value is None:
14601464
continue
1461-
if key.endswith('.ms') or key.endswith('.khz'):
1465+
if key.endswith('.ms'):
14621466
metadata[key] = round(float(value), 3)
1467+
if key.endswith('.hz'):
1468+
metadata[key] = int(round(value))
14631469
if key.endswith('.flag'):
14641470
metadata[key] = bool(value)
14651471
if key.endswith('.y_px/x_px'):
14661472
key_ = key.replace('.y_px/x_px', '.khz/ms')
1467-
metadata[key_] = round(float(value * (y_step_freq / x_step_ms)), 9)
1473+
metadata[key_] = round(float(value * ((y_step_freq / 1000.0) / x_step_ms)), 3)
14681474
metadata.pop(key)
1469-
if key.endswith('.(khz,ms)'):
1475+
if key.endswith('.(hz,ms)'):
14701476
metadata[key] = [
14711477
(
1472-
round(float(val1), 3),
1478+
int(round(val1)),
14731479
round(float(val2), 3),
14741480
)
14751481
for val1, val2 in value
@@ -1488,58 +1494,74 @@ def compute_wrapper(
14881494
if debug_path:
14891495
segments['canvas'].append(canvas[:, trim_begin:trim_end])
14901496

1491-
for key in segments:
1497+
for key in list(segments.keys()):
14921498
value = segments[key]
1493-
segments[key] = np.hstack(value) if len(value) > 0 else stft_db.copy()
1499+
if len(value) == 0:
1500+
segments.pop(key)
1501+
continue
1502+
segments[key] = np.hstack(value)
14941503

14951504
if debug_path:
14961505
cv2.imwrite(join(debug_path, 'spectrogram.tif'), stft_db, [cv2.IMWRITE_TIFF_COMPRESSION, 1])
1497-
cv2.imwrite(
1498-
join(debug_path, 'spectrogram.compressed.tif'),
1499-
segments['stft_db'],
1500-
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1501-
)
15021506
cv2.imwrite(join(debug_path, 'spectrogram.waveplot.png'), waveplot)
1503-
cv2.imwrite(join(debug_path, 'spectrogram.compressed.waveplot.png'), segments['waveplot'])
1504-
temp_top = np.stack((segments['stft_db'], segments['stft_db'], segments['stft_db']), axis=2)
1505-
temp_bot = cv2.resize(
1506-
segments['waveplot'], temp_top.shape[:2][::-1], interpolation=cv2.INTER_LINEAR
1507-
)
1508-
temp_bot = temp_bot.astype(np.float32) * (
1509-
np.iinfo(temp_top.dtype).max / np.iinfo(temp_bot.dtype).max
1510-
)
1511-
temp_bot = np.around(temp_bot).astype(temp_top.dtype)
1512-
temp = np.vstack((temp_top, temp_bot))
1513-
cv2.imwrite(join(debug_path, 'spectrogram.compressed.combined.png'), temp)
1514-
cv2.imwrite(
1515-
join(debug_path, 'spectrogram.compressed.threshold.tif'),
1516-
segments['costs'],
1517-
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1518-
)
1519-
temp = segments['costs'].copy()
1520-
flags = segments['costs'] == 0
1521-
temp = normalize_stft(temp, None, np.uint8)
1522-
temp = cv2.applyColorMap(temp, cv2.COLORMAP_JET)
1523-
temp[:, :, 0][flags] = 0
1524-
temp[:, :, 1][flags] = 0
1525-
temp[:, :, 2][flags] = 0
1526-
cv2.imwrite(
1527-
join(debug_path, 'spectrogram.compressed.threshold.jet.tif'),
1528-
temp,
1529-
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1530-
)
1531-
cv2.imwrite(
1532-
join(debug_path, 'spectrogram.compressed.keypoints.tif'),
1533-
segments['canvas'],
1534-
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1535-
)
1507+
1508+
if 'stft_db' in segments:
1509+
cv2.imwrite(
1510+
join(debug_path, 'spectrogram.compressed.tif'),
1511+
segments['stft_db'],
1512+
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1513+
)
1514+
1515+
if 'waveplot' in segments:
1516+
cv2.imwrite(join(debug_path, 'spectrogram.compressed.waveplot.png'), segments['waveplot'])
1517+
1518+
if 'stft_db' in segments and 'waveplot' in segments:
1519+
temp_top = np.stack((segments['stft_db'], segments['stft_db'], segments['stft_db']), axis=2)
1520+
temp_bot = cv2.resize(
1521+
segments['waveplot'], temp_top.shape[:2][::-1], interpolation=cv2.INTER_LINEAR
1522+
)
1523+
temp_bot = temp_bot.astype(np.float32) * (
1524+
np.iinfo(temp_top.dtype).max / np.iinfo(temp_bot.dtype).max
1525+
)
1526+
temp_bot = np.around(temp_bot).astype(temp_top.dtype)
1527+
temp = np.vstack((temp_top, temp_bot))
1528+
cv2.imwrite(join(debug_path, 'spectrogram.compressed.combined.png'), temp)
1529+
1530+
if 'costs' in segments:
1531+
cv2.imwrite(
1532+
join(debug_path, 'spectrogram.compressed.threshold.tif'),
1533+
segments['costs'],
1534+
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1535+
)
1536+
temp = segments['costs'].copy()
1537+
flags = segments['costs'] == 0
1538+
temp = normalize_stft(temp, None, np.uint8)
1539+
temp = cv2.applyColorMap(temp, cv2.COLORMAP_JET)
1540+
temp[:, :, 0][flags] = 0
1541+
temp[:, :, 1][flags] = 0
1542+
temp[:, :, 2][flags] = 0
1543+
cv2.imwrite(
1544+
join(debug_path, 'spectrogram.compressed.threshold.jet.tif'),
1545+
temp,
1546+
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1547+
)
1548+
1549+
if 'canvas' in segments:
1550+
cv2.imwrite(
1551+
join(debug_path, 'spectrogram.compressed.keypoints.tif'),
1552+
segments['canvas'],
1553+
[cv2.IMWRITE_TIFF_COMPRESSION, 1],
1554+
)
15361555

15371556
output_paths = []
15381557
compressed_paths = []
15391558
datas = [
15401559
(output_paths, 'jpg', stft_db),
1541-
(compressed_paths, 'compressed.jpg', segments['stft_db']),
15421560
]
1561+
if 'stft_db' in segments:
1562+
datas += [
1563+
(compressed_paths, 'compressed.jpg', segments['stft_db']),
1564+
]
15431565

15441566
for accumulator, tag, data in datas:
15451567
if data.dtype != np.uint8:
@@ -1568,29 +1590,31 @@ def compute_wrapper(
15681590
metadata = {
15691591
'wav.path': wav_filepath,
15701592
'spectrogram': {
1571-
'true.path': output_paths,
1593+
'uncompressed.path': output_paths,
15721594
'compressed.path': compressed_paths,
15731595
},
15741596
'threshold.amp': int(round(255.0 * (threshold / max_value))),
1575-
'sr.khz': sr,
1597+
'sr.hz': int(sr),
15761598
'duration.ms': round(duration * 1e3, 3),
15771599
'frequencies': {
1578-
'min.khz': int(FREQ_MIN),
1579-
'max.khz': int(FREQ_MAX),
1580-
'pixels.khz': [round(float(band), 3) for band in bands],
1600+
'min.hz': int(FREQ_MIN),
1601+
'max.hz': int(FREQ_MAX),
1602+
'pixels.hz': bands,
15811603
},
15821604
'size': {
1583-
'true': {
1605+
'uncompressed': {
15841606
'width.px': stft_db.shape[1],
15851607
'height.px': stft_db.shape[0],
15861608
},
1587-
'compressed': {
1588-
'width.px': segments['stft_db'].shape[1],
1589-
'height.px': segments['stft_db'].shape[0],
1590-
},
1609+
'compressed': None,
15911610
},
15921611
'segments': metas,
15931612
}
1613+
if 'stft_db' in segments:
1614+
metadata['size']['compressed'] = {
1615+
'width.px': segments['stft_db'].shape[1],
1616+
'height.px': segments['stft_db'].shape[0],
1617+
},
15941618

15951619
metadata_path = join(output_folder, f'{base}.metadata.json')
15961620
with open(metadata_path, 'w') as metafile:

0 commit comments

Comments
 (0)