fix minor

.ipynb_checkpoints/gradio_artist_classifier-checkpoint.py

@@ -12,6 +12,7 @@ import seaborn as sns
 
 import io
 import json
+import numpy as np
 import skimage.io
 from loguru import logger
 from huggingface_hub import from_pretrained_keras

gradio_artist_classifier.py

@@ -12,6 +12,7 @@ import seaborn as sns
 
 import io
 import json
+import numpy as np
 import skimage.io
 from loguru import logger
 from huggingface_hub import from_pretrained_keras
@@ -25,6 +26,7 @@ from gradcam_utils import get_img_4d_array, make_gradcam_heatmap, align_image_wi
 ARTIST_META = 'artist.json'
 TREND_META = 'trend.json'
 EXAMPLES = ['monet.jpg']
+ALPHA = 0.9
 
 # ---------- Logging ----------
 logger.add('app.log', mode='a')
@@ -68,7 +70,7 @@ def predict(input_image):
                                                             img_4d_array,
                                                             pred_idx=None)
     a_img_pil = align_image_with_heatmap(
-        img_4d_array, a_heatmap, alpha=alpha, cmap='jet')
+        img_4d_array, a_heatmap, alpha=ALPHA, cmap='jet')
     a_img = np.asarray(a_img_pil).astype('float32')/255
     a_label = id2artist[a_pred_id]
     a_prob = a_pred_out[a_pred_id]
@@ -79,7 +81,7 @@ def predict(input_image):
                                                             pred_idx=None)
     
     t_img_pil = align_image_with_heatmap(
-        img_4d_array, t_heatmap, alpha=alpha, cmap='jet')
+        img_4d_array, t_heatmap, alpha=ALPHA, cmap='jet')
     t_img = np.asarray(t_img_pil).astype('float32')/255
     t_label = id2trend[t_pred_id]
     t_prob = t_pred_out[t_pred_id]
@@ -95,7 +97,7 @@ def predict(input_image):
         ax2.imshow(a_img)
         ax3.imshow(t_img)
 
-        ax1.set_title(f'Artist: {artist}\nTrend: {trend}', ha='left', x=0, y=1.05)
+        ax1.set_title(f'Input Image', ha='left', x=0, y=1.05)
         ax2.set_title(f'Artist Prediction:\n =>{a_label} ({a_prob:.2f})', ha='left', x=0, y=1.05)
         ax3.set_title(f'Trend Prediction:\n =>{t_label} ({t_prob:.2f})', ha='left', x=0, y=1.05)
         fig.tight_layout()