report: Touchups in results

Author: jonnor

TODO.md

@@ -7,18 +7,24 @@ Feedback needed
 
 ### Draft 3
 
-Results
 
-- Use Strided-DS-24 as chosen model (confusion matrix etc), instead of auto "best"
-- Finish Discussion and Conclusion
-- Make plots a bit prettier
+### Final 1
 
 Abstract
 
 - Write it!
 - Send to OK for feedback
 
-### Draft 4
+Report
+
+- Add short captions to figures, so List of Figures looks nice
+
+Reprod
+
+- Tag a branch for submitted thesis
+- Upload models to GH
+- Results are from git commit `b49efa5dde48f9fd72a32eff4c751d9d0c0de712`
+- Include perftools Python script in appendix?
 
 Materials
 
@@ -28,24 +34,14 @@ Background
 
 - Make missing images
 
-Report
 
-- Add short captions to figures, so List of Figures looks nice
+### Draft 5
 
 Results
 
 - Include error analysis
 - Plot performance of models relative to fold
 
-Reprod
-
-- Tag a branch for submitted thesis
-- Upload models to GH
-- Results are from git commit `b49efa5dde48f9fd72a32eff4c751d9d0c0de712`
-- Include perftools Python script in appendix?
-
-### Draft 5
-
 Add Acknowledgements?
 
 - Kristian
@@ -59,7 +55,7 @@ Final
 Do they explain the figure setup/contents OK?
 Short captions for the table of contents.
 Remove dots at the end. 2.2, 2.6, 2.16 etc
-- Double check refering style
+- Check references all valid
 - Do a spell checking pass
 - Do a grammar checking pass. LanguageTool + Grammarly
 - Fix Github link, to be to final branch

microesc/report.py

@@ -20,7 +20,7 @@
     'danger': [ 'gun_shot' ],
 }
 
-def plot_confusion(cm, classnames, normalize=False, percent=False):
+def plot_confusion(cm, classnames, normalize=False, percent=False, figsize=(6,5)):
 
     fmt = '.2f'
     if normalize:
@@ -29,14 +29,16 @@ def plot_confusion(cm, classnames, normalize=False, percent=False):
         cm = cm_normalize(cm)*100
         fmt = ".1f"
 
-    fig, ax = plt.subplots(1, figsize=(10,8))
-    seaborn.heatmap(cm, annot=True, ax=ax, fmt=fmt);
+    fig, ax = plt.subplots(1, figsize=figsize)
+    seaborn.heatmap(cm, annot=True, ax=ax, fmt=fmt, cmap='viridis');
 
     ax.set_xlabel('Predicted labels')
     ax.set_ylabel('True labels')
-    ax.set_title('Confusion Matrix') 
     ax.xaxis.set_ticklabels(classnames, rotation=60)
     ax.yaxis.set_ticklabels(classnames, rotation=0)
+
+    fig.tight_layout()
+
     return fig
 
 def cm_normalize(cm):
@@ -309,7 +311,9 @@ def save(fig, name):
 
 
     classnames = urbansound8k.classnames
-    best = df.sort_values('test_acc_mean', ascending=False).head(1).iloc[0]
+    #best = df.sort_values('test_acc_mean', ascending=False).head(1).iloc[0]
+    best = df[df.nickname == 'Stride-DS-24'].iloc[0]
+    print('Selecting', best.nickname)
 
     confusion_matrix = numpy.mean(best.confusions_test, axis=0)
     fig = plot_confusion(confusion_matrix, classnames, percent=True)

report/plot.py

@@ -30,13 +30,13 @@ def model_table(data_path):
 
     table = pandas.DataFrame()
     table['Accuracy (%)'] = df.accuracy*100
-    table['MACC / second'] = [ "{}M".format(int(v/1e6)) for v in df.macc_s ]
-    table['Model parameters'] = [ "{}k".format(int(v/1e3)) for v in df.params ]
+    table['MACC / second'] = [ "{} M".format(int(v/1e6)) for v in df.macc_s ]
+    table['Model parameters'] = [ "{} k".format(int(v/1e3)) for v in df.params ]
     #table['Data augmentation'] = df.augmentation
     table = table.sort_values('Accuracy (%)', ascending=False)
     #table['Time resolution (ms)'] = (df.t_step*1000).astype(int)
 
-    return table.to_latex()
+    return table.to_latex(column_format='lrrr')
 
 def plot_models(data_path, figsize=(12,4), max_params=128e3, max_maccs=4.5e6):
     df = logmel_models(data_path)

report/plots/urbansound8k-existing-models-logmel.tex

@@ -1,4 +1,4 @@
-\begin{tabular}{lrll}
+\begin{tabular}{lrrr}
 \toprule
 {} &  Accuracy (\%) & MACC / second & Model parameters \\
 name         &               &               &                  \\

report/pyincludes/models.py

@@ -46,7 +46,7 @@ def downsample_from_name(name):
     'FLASH': strformat("{:d} kB", (df.flash_usage/1024).astype(int)),
 }, index=df.index)
 
-out = table.to_latex(header=True, index=False)
+out = table.to_latex(header=True, index=False, column_format='lrrllrrr')
 print(out)
 
 outpath = sys.argv[1] 

report/pyincludes/models.tex

@@ -1,4 +1,4 @@
-\begin{tabular}{lllrrlll}
+\begin{tabular}{lrrllrrr}
 \toprule
           Model &   Downsample & Convolution &  L &   F &     MACC &    RAM &   FLASH \\
 \midrule

report/references.bib

@@ -296,7 +296,7 @@ @article{BarcelonaNoiseMonitoring2018
 
 
 
-@article{MAIJALA2018258,
+@article{SourceClassificationSensors,
 title = "Environmental noise monitoring using source classification in sensors",
 journal = "Applied Acoustics",
 volume = "129",
@@ -577,7 +577,6 @@ @article{SONYC2019
     author = "Bello, Juan P. and Silva, Claudio and Nov, Oded and Dubois, R. Luke and Arora, Anish and Salamon, Justin and Mydlarz, Charles and Doraiswamy, Harish",
     doi = "10.1145/3224204",
     title = "SONYC: A System for Monitoring, Analyzing, and Mitigating Urban Noise Pollution",
-    abstract = "Noise is unwanted or harmful sound from environmental sources, including traffic, construction, industrial, and social activity. Noise pollution is one of the topmost quality-of-life concerns for urban residents in the U.S., with more than 70 million people nationwide exposed to noise levels beyond the limit the U.S. Environmental Protection Agency (EPA) considers harmful.12 Such levels have proven effects on health, including sleep disruption, hypertension, heart disease, and hearing loss.5,11,12 In addition, there is evidence of harmful effects on educational performance, with studies showing noise pollution causing learning and cognitive impairment in children, resulting in decreased memory capacity, reading skills, and test scores.",
     number = "2",
     month = "Feb",
     volume = "62",