added table beautification

general_bias_measurement.py

@@ -211,37 +211,40 @@ def evaluate_t2i_model_images(images, prompts, progressBar, debugging, evalType)
     return (sortedDistributionBiasDict, normalisedDistribution, B_D, hallucinationBiases, CLIPMissRates, CLIPErrors)
 def output_eval_results(metrics, topX, evalType):
     sortedDistributionBiasList = list(metrics[0].items())
-    # st.write(list(sortedDistributionBiasDict.values()))
-
-
-    # sortedDistributionBiasList.insert(0, ('object', 'occurrences'))
+    th_props = [
+        ('font-size', '16px'),
+        ('font-weight', 'bold'),
+        ('color', '#ffffff'),
+    ]
+    td_props = [
+        ('font-size', '14px')
+    ]
+
+    styles = [
+        dict(selector="th", props=th_props),
+        dict(selector="td", props=td_props)
+    ]
     col1, col2 = st.columns([0.4,0.6])
     with col1:
         st.write("**Top** "+str(topX-1)+" **Detected Objects**")
-        sortedDistributionBiasList.insert(0, ('object', 'occurrences'))
-        st.table(sortedDistributionBiasList[:topX])
-        # st.write("**Generative Error** $\\varepsilon$")
-        # st.line_chart(sorted(metrics[5], reverse=True))
+        st.table(pd.DataFrame(sortedDistributionBiasList[:topX],
+                              columns=['object', 'occurences'], index=[i+1 for i in range(topX)]
+                              ).style.set_properties().set_table_styles(styles))
+
     with col2:
         st.write("**Distribution of Generated Objects (RAW)** - $B_D$")
         st.bar_chart(metrics[0].values(),color='#1D7AE2')
         st.write("**Distribution of Generated Objects (Normalised)** - $B_D$")
         st.bar_chart(metrics[1],color='#04FB97')
-        # st.write("**Hallucination Bias** - $B_H$")
-        # st.line_chart(sorted(metrics[3], reverse=True))
-        # st.write("**Generative Miss Rate** $M_G$")
-        # st.line_chart(sorted(metrics[4], reverse=True))
+
     if evalType == 'general':
         st.header("\U0001F30E General Bias Evaluation Results")
     else:
         st.header("\U0001F3AF Task-Oriented Bias Evaluation Results")
-    st.table([["Distribution Bias",metrics[2]],["Jaccard Hallucination", np.mean(metrics[3])],
-              ["Generative Miss Rate", np.mean(metrics[4])]])
-    # st.write("Distribution Bias $B_D$ = ", B_D)
-    # st.write("Jaccard Hallucination $H_J$ = ", np.mean(hallucinationBiases))
-    # st.write("Generative Miss Rate $M_G$ = ", np.mean(CLIPMissRates))
-    # st.write("Generative Error $\\varepsilon$ = ", np.mean(CLIPErrors))
-    # progressBar.empty()
+
+    st.table(pd.DataFrame([["Distribution Bias",metrics[2]],["Jaccard Hallucination", np.mean(metrics[3])],
+                          ["Generative Miss Rate", np.mean(metrics[4])]],
+                          columns=['metric','value'], index=[' ' for i in range(3)]))
 def BLIP_captioning_single(image, gen_kwargs):
     caption = None
     inputs = BLIP_processor(image, return_tensors="pt").to(device)