def main():
_CLUSTERING_TYPE = 'n2d'
_BATCH_SIZE = 496
_N_BATCH = 5
_PLOTS = True
_MOSAICS = True
_SILHOUETTE = True
_OUTPUT_IMAGE_WIDTH = 96
_OUTPUT_IMAGE_HEIGHT = 54
_MOSAIC_NROW = 10
_MOSAIC_NCOL_MAX = 10
model, model_config = Tools.load_model(
CONFIG,
CONFIG.get('clustering')[_CLUSTERING_TYPE]['model']['type'],
CONFIG.get('clustering')[_CLUSTERING_TYPE]['model']['name'])
filenames = Tools.list_directory_filenames(
CONFIG.get('sampling')['autoencoder']['directory']['train'])
pictures_id, intermediate_output = Tools.load_latent_representation(
CONFIG, model, model_config, filenames, _BATCH_SIZE, _N_BATCH, False)
clustering = run_clustering(CONFIG, _CLUSTERING_TYPE, pictures_id,
intermediate_output)
clustering.save()
if _PLOTS:
run_plots(CONFIG, _CLUSTERING_TYPE, clustering)
if _SILHOUETTE:
plot_silhouette(CONFIG, _CLUSTERING_TYPE, clustering)
if _MOSAICS:
plot_mosaics(CONFIG, _CLUSTERING_TYPE, clustering, _OUTPUT_IMAGE_WIDTH,
_OUTPUT_IMAGE_HEIGHT, _MOSAIC_NROW, _MOSAIC_NCOL_MAX)
def populate_embedding(config, db_manager, clustering_type, clustering_version, clustering_model_type, clustering_model_name, drop=False):
clustering, clustering_config = Tools.load_clustering(CONFIG, clustering_type, clustering_version, clustering_model_type, clustering_model_name)
if drop:
db_manager.drop_embedding_partition(clustering_type, clustering_version, clustering_model_type, clustering_model_name)
if clustering_type == 'n2d':
clustering = N2DClustering(clustering_config)
elif clustering_type == 'classical':
clustering = ClassicalClustering(clustering_config)
elif clustering_type == 'dbscan':
clustering = DBScanClustering(clustering_config)
else:
raise Exception
clustering.load()
model, model_config = Tools.load_model(CONFIG, clustering_model_type, clustering_model_name)
filenames = Tools.list_directory_filenames(CONFIG.get('directory')['collections'])
generator = Tools.load_latent_representation(CONFIG, model, model_config, filenames, 496, None, True)
count = 0
for ids, latents in generator:
pictures_embedding = clustering.predict_embedding(latents)
pictures_label = clustering.predict_label(pictures_embedding)
rows = []
for i, id in enumerate(ids):
rows.append((
id,
float(np.round(pictures_embedding[i][0], 4)),
float(np.round(pictures_embedding[i][1], 4)),
int(pictures_label[i]),
clustering_type,
clustering_version,
clustering_model_type,
clustering_model_name
))
count += db_manager.insert_row_pictures_embedding(rows)
print("Nombre d'insertion: %s / %s" % (count, len(filenames)))
return
