From f574c1bf8e813bbe3da2eb8162c072fa69b51c57 Mon Sep 17 00:00:00 2001
From: "Daniel.Frisinghelli" <daniel.frisinghelli@eurac.edu>
Date: Thu, 18 Feb 2021 16:38:06 +0100
Subject: [PATCH] Adjust counters starting from 0.
---
pysegcnn/core/trainer.py | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/pysegcnn/core/trainer.py b/pysegcnn/core/trainer.py
index 4e82d54..bc0f11c 100644
--- a/pysegcnn/core/trainer.py
+++ b/pysegcnn/core/trainer.py
@@ -2545,24 +2545,28 @@ class NetworkInference(BaseConfig):
# check whether to reconstruct the scenes of a dataset
if self.predict_scene:
+ # append model predictions of current batch to scene dictionary
+ for k, v in zip(INFERENCE_NAMES, [inputs, labels, prdctn]):
+ scenes[k].append(v)
+
# check if an entire scene is processed
- if batch % self.trg_ds.dataset.tiles == 0 and batch != 0:
+ if (batch + 1) % self.trg_ds.dataset.tiles == 0 and batch != 0:
# convert scene dictionary to numpy arrays
- inputs, labels, prdctn = [np.asarray(v) for _, v in
- scenes.items()]
+ inputs, labels, prdctn = [
+ np.asarray(v) for _, v in scenes.items()]
# tiles of the current scene
current_tiles = self.trg_ds.indices[
- np.arange(batch - self.trg_ds.dataset.tiles, batch)]
+ np.arange((batch + 1) - self.trg_ds.dataset.tiles,
+ batch)]
# name of the current scene
batch = np.unique([self.trg_ds.dataset.scenes[sid]['id']
for sid in current_tiles]).item()
# modify the progress string
- progress = progress.replace('Sample', 'Scene')
- progress += ' Id: {}'.format(batch)
+ progress = 'Scene Id: {}'.format(batch)
# reconstruct the entire scene
inputs = reconstruct_scene(inputs)
@@ -2603,10 +2607,6 @@ class NetworkInference(BaseConfig):
self.scenes_path.joinpath(batch_name)),
bbox_inches='tight')
- # append model predictions of current batch to scene dictionary
- for k, v in zip(INFERENCE_NAMES, [inputs, labels, prdctn]):
- scenes[k].append(v)
-
else:
# save current batch to output dictionary
output[batch] = {k: v for k, v in zip(INFERENCE_NAMES,
--
GitLab