Prepare for release 0.2.0

Author: Temigo

CHANGES.txt

@@ -1 +1,3 @@
 v0.1.1, 3/30/2018 -- Initial release.
+
+v0.2.0, 4/9/2018 -- 3D version of PPN initial release.

README.md

@@ -1,10 +1,11 @@
 # faster-particles
 This package includes the following:
 * Toydata generator
-* LArCV data interface
+* LArCV data interface (2D and 3D)
 * Pixel Proposal Network implementation using Tensorflow
 
 ## Installation
+### Dependencies
 You must install [larcv2](https://github.com/DeepLearnPhysics/larcv2) and its
 own dependencies (ROOT, OpenCV, Numpy) in order to use LArCV data interface.
 To install `larcv2`:
@@ -15,6 +16,7 @@ source configure.sh
 make
 ```
 
+### Install
 Then install `faster-particles` with Pip:
 ```bash
 pip install faster-particles
@@ -35,22 +37,29 @@ To use toydata rather than LArCV data in the following sections, use option `--t
 LArCV data files can be specified with `--data` option. They can use regex, e.g. `ppn_p[01]*.root`.
 
 ### Training
-To train PPN on 1000 steps:
+The program output is divided between:
+* Output directory: with all the weights
+* Log directory: to store all Tensorflow logs (and visualize them with Tensorboard)
+* Display directory: stores regular snapshots taken during training of PPN1 and PPN2 proposals compared to ground truth.
+
+To train PPN on 1000 steps use:
 ```bash
 ppn train -o output/dir/ -l log/dir/ -d display/dir -n ppn -m 1000 --data path/to/data
 ```
 
-To train the base network (currently only VGG available) on track/shower classification task:
+To train the base network (currently only VGG available) on track/shower classification task use:
 ```bash
 ppn train -o output/dir/ -l log/dir/ -d display/dir -n base -m 1000
 ```
 
+To train on 3D data, use the argument `-3d` and don't forget to specify the image size with `-N` argument (e.g. 192 for a compression factor of 4, see `larcvdata_generator.py` for more details).
 
 ### Inference
 To run inference with a minimal score of 0.5 for predicted points:
 ```bash
 ppn demo weights_file.ckpt -d display/dir/ -ms 0.5
 ```
+The display directory will contain snapshots of the results.
 
 More options are available through `ppn train -h` and `ppn demo -h` respectively.
 

faster_particles/config.py

@@ -32,7 +32,8 @@ class PPNConfig(object):
     TOYDATA = False
     # DATA = "/data/drinkingkazu/dlprod_ppn_v05/ppn_p[01]*.root"
     # DATA = "/stage/drinkingkazu/dlprod_ppn_v05/ppn_p01.root" # For 2D
-    DATA = "/data/drinkingkazu/dlprod_ppn_v05/ppn_p00_0000_0019.root" # For 3D
+    # DATA = "/data/drinkingkazu/dlprod_ppn_v05/ppn_p00_0000_0019.root" # For 3D
+    DATA = ""
     DATA_3D = False
 
     # Track configuration

faster_particles/demo_ppn.py

@@ -239,8 +239,6 @@ def closest_gt_distance(im_proposals, gt_pixels):
 
 def distances_plot(cfg, distances):
     plt.gcf().clear()
-    print(distances)
-    print(np.count_nonzero(np.greater(distances, 5)), len(distances))
     bins = np.linspace(0, 5, 20)
     plt.hist(distances, 100)
     plt.xlabel("distance to closest ground truth pixel")
@@ -288,8 +286,6 @@ def inference(cfg):
             blob = data.forward()
             summary, results = net.test_image(sess, blob)
             if cfg.NET == 'ppn':
-                #print(blob, results)
-                #im_proposals.extend(results['im_proposals'])
                 im_labels.extend(results['im_labels'])
                 im_scores.extend(results['im_scores'])
                 im_proposals_filtered = display(blob, cfg, index=i, dim1=net.dim1, dim2=net.dim2, **results)

requirements.txt

@@ -1,5 +1,5 @@
-matplotlib
-numpy
-scikit-learn
-scikit-image
-tensorflow-gpu
+matplotlib>=2.2.2
+numpy>=1.14.2
+scikit-learn>=0.18.1
+scikit-image>=0.12.3
+tensorflow>=1.3.1

setup.py

@@ -3,7 +3,7 @@
 
 setup(
     name='faster-particles',
-    version='0.1.6',
+    version='0.2.0',
     author='Laura Domine, Ji Won Park, Kazuhiro Terao',
     author_email='temigo@gmx.com',
     packages=find_packages(),
@@ -18,7 +18,7 @@
     long_description_content_type='text/markdown',
     url='https://github.com/Temigo/faster-particles',
     install_requires=[
-        "matplotlib >= 1.5.3",
+        "matplotlib >= 2.2.2",
         "numpy >= 1.13.1",
         "scikit-learn >= 0.18.1",
         "scikit-image >= 0.12.3",