diff options
| -rw-r--r-- | cli/app/commands/biggan/extract_dense_vectors.py | 2 | ||||
| -rw-r--r-- | cli/app/search/search_class.py | 5 |
2 files changed, 4 insertions, 3 deletions
diff --git a/cli/app/commands/biggan/extract_dense_vectors.py b/cli/app/commands/biggan/extract_dense_vectors.py index f7cffb2..3ca7313 100644 --- a/cli/app/commands/biggan/extract_dense_vectors.py +++ b/cli/app/commands/biggan/extract_dense_vectors.py @@ -24,7 +24,7 @@ from app.search.params import timestamp help='Normalize labels every N steps') @click.option('-feat', '--use_feature_detector', 'opt_use_feature_detector', is_flag=True, help='Compute feature loss') -@click.option('-ll', '--feature_layers', 'opt_feature_layers', default="1a,2a,3a,4a,7a", +@click.option('-ll', '--feature_layers', 'opt_feature_layers', default="1a,3a,7a", help='Feature layers used for loss') @click.option('-snap', '--snapshot_interval', 'opt_snapshot_interval', default=20, help='Interval to store sample images') diff --git a/cli/app/search/search_class.py b/cli/app/search/search_class.py index 921e586..ccb10cd 100644 --- a/cli/app/search/search_class.py +++ b/cli/app/search/search_class.py @@ -113,7 +113,7 @@ def find_nearest_vector(sess, generator, opt_fp_in, opt_dims, out_images, out_la ## normalize the Y encoding # normalized_labels = tf.nn.l2_normalize(input_y) # tf.reduce_mean(tf.abs(encoding - gen_encoding)) - normalized_labels = input_y / tf.reduce_sum(input_y) + normalized_labels = input_y / tf.reduce_max(input_y) normalized_alpha = tf.compat.v1.placeholder(dtype=np.float32, shape=()) clip_labels = tf.assign(input_y, normalized_labels * (1 - normalized_alpha) + input_y * normalized_alpha) @@ -196,7 +196,8 @@ def find_nearest_vector(sess, generator, opt_fp_in, opt_dims, out_images, out_la if opt_stochastic_clipping and (i % opt_clip_interval) == 0: # and i < opt_steps * 0.45: sess.run(clip_latent, { clipped_alpha: (i / opt_steps) * 2 }) if opt_label_clipping and (i % opt_clip_interval) == 0: # and i < opt_steps * 0.75: - sess.run(clip_labels, { normalized_alpha: (i / opt_steps) ** 2 }) + # sess.run(clip_labels, { normalized_alpha: (i / opt_steps) ** 2 }) + sess.run(clip_labels, { normalized_alpha: i / opt_steps }) if opt_video and opt_snapshot_interval != 0 and (i % opt_snapshot_interval) == 0: phi_guess = sess.run(output) guess_im = imgrid(imconvert_uint8(phi_guess), cols=1) |