Updated the download functionality

For web-app purposes, a download button is necessary as saving locally doesn't make sense. Just a simple

tab_manager.py

@@ -90,20 +90,21 @@ def general_bias_eval_setup(tab, modelID, imagesTab):
 
         if user_evaluation_variables.RUN_TIME and user_evaluation_variables.CURRENT_EVAL_TYPE == 'general':
             GBM.output_eval_results(user_evaluation_variables.EVAL_METRICS, 21, 'general')
+            genCSVData = create_word_distribution_csv(user_evaluation_variables.EVAL_METRICS,
+                                                   user_evaluation_variables.EVAL_ID,
+                                                   'general')
             st.write("\U0001F553 Time Taken: ", user_evaluation_variables.RUN_TIME)
-
             saveEvalsButton = st.button("Save + Upload Evaluations", key='SAVE_EVAL_GEN')
-            saveDistButton = st.button("Download Object Distribution", key='SAVE_TOP_GEN')
+            st.download_button(label="Download Object Distribution data", data=genCSVData, key='SAVE_TOP_GEN',
+                               file_name=user_evaluation_variables.EVAL_ID + '_general' + '_word_distribution.csv',
+                               mime='text/csv')
+
             if saveEvalsButton:
                 st.write("Saving and uploading evaluations")
                 user_evaluation_variables.update_evaluation_table('general',False)
                 user_evaluation_variables.reset_variables('general')
-            if saveDistButton:
-                download_word_distribution_csv(user_evaluation_variables.EVAL_METRICS,
-                                               user_evaluation_variables.EVAL_ID, 'general')
-
 
-def task_oriented_bias_eval_setup(tab,modelID,imagesTab):
+def task_oriented_bias_eval_setup(tab, modelID, imagesTab):
     biasSetupDF_EVAL = pd.DataFrame(
         {
             "TO Eval. Variable": ["No. Images to Generate per prompt", "No. Inference Steps", "Image Size (N x N)"],
@@ -144,33 +145,31 @@ def task_oriented_bias_eval_setup(tab,modelID,imagesTab):
             # update_images_tab(imagesTab)
         if user_evaluation_variables.RUN_TIME and user_evaluation_variables.CURRENT_EVAL_TYPE == 'task-oriented':
             GBM.output_eval_results(user_evaluation_variables.EVAL_METRICS, 21, 'task-oriented')
+            taskCSVData = create_word_distribution_csv(user_evaluation_variables.EVAL_METRICS,
+                                                   user_evaluation_variables.EVAL_ID,
+                                                   user_evaluation_variables.TASK_TARGET)
             st.write("\U0001F553 Time Taken: ", user_evaluation_variables.RUN_TIME)
             saveEvalsButton = st.button("Save + Upload Evaluations", key='SAVE_EVAL_TASK')
-            saveDistButton = st.button("Download Object Distribution", key='SAVE_TOP_TASK')
+            st.download_button(label="Download Object Distribution data", data=taskCSVData, key='SAVE_TOP_TASK',
+                               file_name=user_evaluation_variables.EVAL_ID+'_'+user_evaluation_variables.TASK_TARGET+'_word_distribution.csv',
+                               mime='text/csv')
+
             if saveEvalsButton:
                 st.write("Saving and uploading evaluations")
-                user_evaluation_variables.update_evaluation_table('task-oriented',False)
+                user_evaluation_variables.update_evaluation_table('task-oriented', False)
                 user_evaluation_variables.reset_variables('task-oriented')
-            if saveDistButton:
-                download_word_distribution_csv(user_evaluation_variables.EVAL_METRICS,
-                                               user_evaluation_variables.EVAL_ID, user_evaluation_variables.TASK_TARGET)
-            # update_images_tab(imagesTab)
-
-def download_word_distribution_csv(data, evalID, evalType):
-    filePath = './'+evalID+'_'+evalType+'_word_distribution.csv'
 
+def create_word_distribution_csv(data, evalID, evalType):
     listOfObjects = list(data[0].items())
-    with open(filePath, 'w', newline='') as fp:
-        csvwriter = csv.writer(fp)
-        csvwriter.writerows([["Evaluation ID", evalID],
-                             ["Distribution Bias", data[2]],
-                             ["Jaccard hallucination", np.mean(data[3])],
-                             ["Generative Miss Rate", np.mean(data[4])]])
-        csvwriter.writerow(['Position', 'Object', 'No. Occurences', 'Normalized'])
-        for obj, val, norm, ii in zip(listOfObjects, data[0].values(), data[1], range(len(listOfObjects))):
-            csvwriter.writerow([ii, obj[0], val, norm])
-    st.success('Successfully downloaded word distribution data!', icon="✅")
-
+    csvContents = [["Evaluation Type/Target", evalType],
+                   ["Evaluation ID", evalID],
+                   ["Distribution Bias", data[2]],
+                   ["Jaccard hallucination", np.mean(data[3])],
+                   ["Generative Miss Rate", np.mean(data[4])],
+                   ['Position', 'Object', 'No. Occurences', 'Normalized']]
+    for obj, val, norm, ii in zip(listOfObjects, data[0].values(), data[1], range(len(listOfObjects))):
+        csvContents.append([ii, obj[0], val, norm])
+    return pd.DataFrame(csvContents).to_csv(header=False,index=False).encode('utf-8')
 def initiate_general_bias_evaluation(tab, modelID, specs, imagesTab):
     startTime = time.time()
     objectData = None
@@ -202,9 +201,9 @@ def initiate_general_bias_evaluation(tab, modelID, specs, imagesTab):
             st.write(" ***Occupations*** = ", specs[1]["Check"][2])
         st.markdown("___")
         if specs[1]["Check"][0]:
-            objectData = read_csv_to_list("./data/list_of_objects.csv")
+            objectData = read_csv_to_list("./list_of_objects.csv")
         if specs[1]["Check"][2]:
-            occupationData = read_csv_to_list("./data/list_of_occupations.csv")
+            occupationData = read_csv_to_list("./list_of_occupations.csv")
         if objectData == None and occupationData == None:
             st.error('Make sure that at least one of the "Objects" or "Occupations" rows are checked', icon="🚨")
         else:
@@ -250,7 +249,6 @@ def initiate_general_bias_evaluation(tab, modelID, specs, imagesTab):
             user_evaluation_variables.EVAL_METRICS = GBM.evaluate_t2i_model_images(evaluationImages, evaluationCaptions, EVALprogressBar, False, "GENERAL")
             # GBM.output_eval_results(user_evaluation_variables.EVAL_METRICS, 21)
             elapsedTime = time.time() - startTime
-            # st.write("\U0001F553 Time Taken: ", str(datetime.timedelta(seconds=elapsedTime)).split(".")[0])
 
             user_evaluation_variables.NO_SAMPLES = len(evaluationImages)
             user_evaluation_variables.RESOLUTION = specs[0]["GEN Values"][2] + "x" + specs[0]["GEN Values"][2]
@@ -294,7 +292,7 @@ def initiate_task_oriented_bias_evaluation(tab, modelID, specs, target, imagesTa
                      icon="🚨")
         else:
             COCOLoadingBar = st.progress(0, text="Scanning through COCO Dataset for relevant prompts. Please wait")
-            prompts, cocoIDs = get_COCO_captions('./data/COCO_captions.json', target.lower(), COCOLoadingBar, captionsToExtract)
+            prompts, cocoIDs = get_COCO_captions('./COCO_captions.json', target.lower(), COCOLoadingBar, captionsToExtract)
             if len(prompts) == 0:
                 st.error('Woops! Could not find **ANY** relevant COCO prompts for the target: '+target.lower()+
                          '\nPlease input a different target', icon="🚨")
@@ -318,10 +316,7 @@ def initiate_task_oriented_bias_evaluation(tab, modelID, specs, target, imagesTa
                 EVALprogressBar = st.progress(0, text="Evaluating " + modelID + " Model Images. Please wait.")
                 user_evaluation_variables.EVAL_METRICS = GBM.evaluate_t2i_model_images(TASKImages, TASKCaptions[0], EVALprogressBar, False, "TASK")
 
-
-                # GBM.output_eval_results(user_evaluation_variables.EVAL_METRICS, 21)
                 elapsedTime = time.time() - startTime
-                # st.write("\U0001F553 Time Taken: ", str(datetime.timedelta(seconds=elapsedTime)).split(".")[0])
 
                 user_evaluation_variables.NO_SAMPLES = len(TASKImages)
                 user_evaluation_variables.RESOLUTION = specs["TO Values"][2]+"x"+specs["TO Values"][2]
@@ -340,16 +335,19 @@ def initiate_task_oriented_bias_evaluation(tab, modelID, specs, target, imagesTa
                 user_evaluation_variables.TASK_COCOIDs = cocoIDs
 
                 user_evaluation_variables.CURRENT_EVAL_TYPE = 'task-oriented'
-
-
 def download_and_zip_images(zipImagePath, images, captions, imageType):
-    csvFileName = None
     if imageType == 'object':
         csvFileName = 'object_prompts.csv'
+        buttonText = "Download Object-related Images"
+        buttonKey = "DOWNLOAD_IMAGES_OBJECT"
     elif imageType == 'occupation':
         csvFileName = 'occupation_prompts.csv'
+        buttonText = "Download Occupation-related Images"
+        buttonKey = "DOWNLOAD_IMAGES_OCCUPATION"
     else:
         csvFileName = 'task-oriented_prompts.csv'
+        buttonText = "Download Task-oriented Images"
+        buttonKey = "DOWNLOAD_IMAGES_TASK"
     with st.spinner("Zipping images..."):
         with zipfile.ZipFile(zipImagePath, 'w') as img_zip:
             for idx, image in enumerate(images):
@@ -372,7 +370,9 @@ def download_and_zip_images(zipImagePath, images, captions, imageType):
                     csvwriter.writerow([id, prompt])
 
             img_zip.writestr(csvFileName, string_buffer.getvalue())
-    st.success('Successfully zipped and downloaded images!', icon="✅")
+        with open(zipImagePath, 'rb') as f:
+            st.download_button(label=buttonText, data=f, key=buttonKey,
+                               file_name=zipImagePath)
 
 
 def update_images_tab(imagesTab):
@@ -395,11 +395,9 @@ def update_images_tab(imagesTab):
                 for idx, image in enumerate(user_evaluation_variables.OBJECT_IMAGES):
                     next(cols).image(image, width=225, caption=user_evaluation_variables.OBJECT_CAPTIONS[1][idx])
 
-            saveObjectImages = st.button("Save Object-related Images")
-            if saveObjectImages:
-                zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.EVAL_ID + '_object_related_images.zip'
-                download_and_zip_images(zipPath, user_evaluation_variables.OBJECT_IMAGES,
-                                        user_evaluation_variables.OBJECT_CAPTIONS, 'object')
+            zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.DATE + '_' + user_evaluation_variables.TIME + '_object_related_images.zip'
+            download_and_zip_images(zipPath, user_evaluation_variables.OBJECT_IMAGES,
+                                    user_evaluation_variables.OBJECT_CAPTIONS, 'object')
 
         if len(user_evaluation_variables.OCCUPATION_IMAGES) > 0:
             user_evaluation_variables.OCCUPATION_IMAGES_IN_UI = True
@@ -419,11 +417,10 @@ def update_images_tab(imagesTab):
                 for idx, image in enumerate(user_evaluation_variables.OCCUPATION_IMAGES):
                     next(cols).image(image, width=225, caption=user_evaluation_variables.OCCUPATION_CAPTIONS[1][idx])
 
-            saveOccupationImages = st.button("Save Occupation-related Images")
-            if saveOccupationImages:
-                zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.EVAL_ID + '_occupation_related_images.zip'
-                download_and_zip_images(zipPath, user_evaluation_variables.OCCUPATION_IMAGES,
-                                        user_evaluation_variables.OCCUPATION_CAPTIONS, 'occupation')
+            zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.DATE + '_' + user_evaluation_variables.TIME + '_occupation_related_images.zip'
+
+            download_and_zip_images(zipPath, user_evaluation_variables.OCCUPATION_IMAGES,
+                                    user_evaluation_variables.OCCUPATION_CAPTIONS, 'occupation')
         if len(user_evaluation_variables.TASK_IMAGES) > 0:
             with st.expander(user_evaluation_variables.TASK_TARGET+'-related Images'):
                 user_evaluation_variables.TASK_IMAGES_IN_UI = True
@@ -443,11 +440,10 @@ def update_images_tab(imagesTab):
                 for idx, image in enumerate(user_evaluation_variables.TASK_IMAGES):
                     next(cols).image(image, width=225, caption=user_evaluation_variables.TASK_CAPTIONS[1][idx])
 
-            saveTaskImages = st.button("Save Task-oriented Images")
-            if saveTaskImages:
-                zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.EVAL_ID + '_'+ user_evaluation_variables.TASK_TARGET+'-oriented_images.zip'
-                download_and_zip_images(zipPath, user_evaluation_variables.TASK_IMAGES,
-                                        user_evaluation_variables.TASK_CAPTIONS, 'task-oriented')
+            zipPath = 'TBYB_' + user_evaluation_variables.USERNAME + '_' + user_evaluation_variables.DATE + '_' + user_evaluation_variables.TIME + '_' + user_evaluation_variables.TASK_TARGET + '_related_images.zip'
+            download_and_zip_images(zipPath, user_evaluation_variables.TASK_IMAGES,
+                                    user_evaluation_variables.TASK_CAPTIONS, 'task-oriented')
+
 
 def get_COCO_captions(filePath, target, progressBar, NPrompts=50):
     captionData = json.load(open(filePath))
@@ -455,12 +451,12 @@ def get_COCO_captions(filePath, target, progressBar, NPrompts=50):
     COCOIDs = []
     random.seed(42)
     random.shuffle(captionData['annotations'])
-    for anno in captionData['annotations']:
+    for anno, pp in zip(captionData['annotations'], range(len(captionData['annotations']))):
         if target in anno.get('caption').lower().split(' '):
             if len(COCOCaptions) < NPrompts:
                 COCOCaptions.append(anno.get('caption').lower())
                 COCOIDs.append(str(anno.get('id')))
-        percentComplete = len(COCOCaptions) / NPrompts
+        percentComplete = pp/len(captionData['annotations'])
         progressBar.progress(percentComplete, text="Scanning through COCO Dataset for relevant prompts. Please wait")
     return (COCOCaptions, COCOIDs)
 def read_csv_to_list(filePath):