Error: {e} Check the server log for more information.' - - -@app.route('/ocrpage', methods=['POST']) -def ocrpage(): - data = request.json - try: - return utils.render_ocr(data["mmif_id"], data['vid_path'], data["view_id"], data["page_number"]) - except Exception as e: - return f'
Unexpected error of type {type(e)}: {e}' + if "page_number" not in request.json: + build_ocr_tab(request.json) + request.json["page_number"] = 0 + # return serve_first_ocr_page(request.json) + # else: + return serve_ocr_page(request.json) @app.route('/upload', methods=['GET', 'POST']) @@ -93,7 +85,7 @@ def display(viz_id): return html_file else: app.logger.debug(f"Visualization {viz_id} not found in cache.") - os.remove(path) + rmtree(path) flash("File not found -- please upload again (it may have been deleted to clear up cache space).") return redirect("/upload") @@ -103,6 +95,45 @@ def send_js(path): return send_from_directory("uv", path) +def render_mmif(mmif_str, viz_id): + mmif = Mmif(mmif_str) + rendered_documents = render_documents(mmif, viz_id) + rendered_annotations = render_annotations(mmif, viz_id) + return render_template('player.html', + docs=rendered_documents, + viz_id=viz_id, + annotations=rendered_annotations) + + +def build_ocr_tab(data): + """ + Prepares OCR (at load time, due to lazy loading) + """ + try: + data = dict(request.json) + mmif_str = open(cache.get_cache_root() / + data["mmif_id"] / "file.mmif").read() + mmif = Mmif(mmif_str) + ocr_view = mmif.get_view_by_id(data["view_id"]) + prepare_ocr(mmif, ocr_view, data["mmif_id"]) + request.json["vid_path"] = mmif.get_documents_by_type(DocumentTypes.VideoDocument)[ + 0].location_path() + + except Exception as e: + app.logger.error(f"{e}\n{traceback.format_exc()}") + return f'
Error: {e} Check the server log for more information.' + + +def serve_ocr_page(data): + """ + Serves subsequent OCR pages + """ + try: + return render_ocr_page(data["mmif_id"], data['vid_path'], data["view_id"], data["page_number"]) + except Exception as e: + return f'
Unexpected error of type {type(e)}: {e}'
+
+
def upload_file(in_mmif):
# Save file locally
in_mmif_bytes = in_mmif if isinstance(in_mmif, bytes) else in_mmif.read()
@@ -117,13 +148,7 @@ def upload_file(in_mmif):
with open(path / 'file.mmif', 'w') as in_mmif_file:
app.logger.debug(f"Writing original MMIF to {path / 'file.mmif'}")
in_mmif_file.write(in_mmif_str)
- mmif = Mmif(in_mmif_str)
- htmlized_docs = utils.documents_to_htmls(mmif, viz_id)
- app.logger.debug(f"Prepared document: {[d[0] for d in htmlized_docs]}")
- annotations = utils.prep_annotations(mmif, viz_id)
- app.logger.debug(f"Prepared Annotations: {[annotation[0] for annotation in annotations]}")
- html_page = render_template('player.html',
- docs=htmlized_docs, viz_id=viz_id, annotations=annotations)
+ html_page = render_mmif(in_mmif_str, viz_id)
with open(os.path.join(path, "index.html"), "w") as f:
f.write(html_page)
except FileExistsError:
@@ -133,7 +158,6 @@ def upload_file(in_mmif):
t = Thread(target=cleanup)
t.daemon = True
t.run()
-
agent = request.headers.get('User-Agent')
if 'curl' in agent.lower():
return f"Visualization ID is {viz_id}\nYou can access the visualized file at {request.url_root}display/{viz_id}\n"
@@ -143,7 +167,8 @@ def upload_file(in_mmif):
if __name__ == '__main__':
# Make path for temp files
cache_path = cache.get_cache_root()
- cache_symlink_path = os.path.join(app.static_folder, cache._CACHE_DIR_SUFFIX)
+ cache_symlink_path = os.path.join(
+ app.static_folder, cache._CACHE_DIR_SUFFIX)
if os.path.islink(cache_symlink_path):
os.unlink(cache_symlink_path)
elif os.path.exists(cache_symlink_path):
@@ -158,5 +183,5 @@ def upload_file(in_mmif):
port = 5000
if len(sys.argv) > 2 and sys.argv[1] == '-p':
port = int(sys.argv[2])
-
+
app.run(port=port, host='0.0.0.0', debug=True, use_reloader=True)
diff --git a/displacy/__init__.py b/displacy/__init__.py
index 0ae0ffe..5a76760 100644
--- a/displacy/__init__.py
+++ b/displacy/__init__.py
@@ -48,11 +48,11 @@ def read_text(textdoc, app_root):
# container, see the comment in html_text() in ../app.py)
if not os.path.isfile(location):
if location.startswith('file:///'):
- location = location[8:]
+ location = location[7:]
else:
# this should not happen anymore, but keeping it anyway
location = location[1:]
- location = os.path.join(app_root, 'static', location)
+ # location = os.path.join(app_root, 'static', location)
with open(location) as fh:
text = fh.read()
else:
diff --git a/examples/whisper-spacy.json b/examples/whisper-spacy.json
index 967a3d4..9a164e6 100644
--- a/examples/whisper-spacy.json
+++ b/examples/whisper-spacy.json
@@ -8,7 +8,7 @@
"properties": {
"mime": "video",
"id": "d1",
- "location": "file:///data/video/service-mbrs-ntscrm-01181182.mp4"
+ "location": "file:///data/service-mbrs-ntscrm-01181182.mp4"
}
},
{
@@ -16,7 +16,7 @@
"properties": {
"mime": "audio",
"id": "d2",
- "location": "file:///data/audio/service-mbrs-ntscrm-01181182.wav"
+ "location": "file:///data/service-mbrs-ntscrm-01181182.wav"
}
},
{
@@ -24,7 +24,7 @@
"properties": {
"mime": "text",
"id": "d3",
- "location": "file:///data/text/service-mbrs-ntscrm-01181182.txt"
+ "location": "file:///data/service-mbrs-ntscrm-01181182.txt"
}
}
],
diff --git a/ocr.py b/ocr.py
index a964296..dc1bbc6 100644
--- a/ocr.py
+++ b/ocr.py
@@ -1,14 +1,13 @@
import datetime
-import pathlib
import cv2
-import tempfile
import json
import re
-import os, shutil
+import os
+import shutil
+from mmif.vocabulary.annotation_types import AnnotationTypes
+from mmif.vocabulary.document_types import DocumentTypes
-from flask import render_template
-from mmif import AnnotationTypes, DocumentTypes, Mmif
from mmif.utils.video_document_helper import convert_timepoint, convert_timeframe
import cache
@@ -51,14 +50,17 @@ def update(self, anno, mmif):
elif anno.at_type.shortname == "Paragraph":
view = mmif.get_view_by_id(anno.parent)
- text_anno = mmif[anno.properties.get("document")]
+ text_anno = view.get_annotation_by_id(
+ anno.properties.get("document"))
self.add_text_document(text_anno)
- def add_bounding_box(self, anno, mmif: Mmif):
- timepoint_anno = None
+ def add_bounding_box(self, anno, mmif):
if "timePoint" in anno.properties:
timepoint_anno = mmif[anno.get("timePoint")]
-
+
+ if timepoint_anno:
+ self.add_timepoint(timepoint_anno, mmif,
+ skip_if_view_has_frames=False)
else:
for alignment_anns in mmif.get_alignments(AnnotationTypes.BoundingBox, AnnotationTypes.TimePoint).values():
for alignment_ann in alignment_anns:
@@ -88,9 +90,11 @@ def add_bounding_box(self, anno, mmif: Mmif):
def add_timeframe(self, anno, mmif):
# If annotation has multiple targets, pick the first and last as start and end
if "targets" in anno.properties:
- start_id, end_id = anno.properties.get("targets")[0], anno.properties.get("targets")[-1]
+ start_id, end_id = anno.properties.get(
+ "targets")[0], anno.properties.get("targets")[-1]
anno_parent = mmif.get_view_by_id(anno.parent)
- start_anno, end_anno = mmif[start_id], mmif[end_id]
+ start_anno, end_anno = anno_parent.get_annotation_by_id(
+ start_id), anno_parent.get_annotation_by_id(end_id)
start = convert_timepoint(mmif, start_anno, "frames")
end = convert_timepoint(mmif, end_anno, "frames")
start_secs = convert_timepoint(mmif, start_anno, "seconds")
@@ -99,7 +103,8 @@ def add_timeframe(self, anno, mmif):
start, end = convert_timeframe(mmif, anno, "frames")
start_secs, end_secs = convert_timeframe(mmif, anno, "seconds")
self.range = (start, end)
- self.timestamp_range = (str(datetime.timedelta(seconds=start_secs)), str(datetime.timedelta(seconds=end_secs)))
+ self.timestamp_range = (str(datetime.timedelta(seconds=start_secs)), str(
+ datetime.timedelta(seconds=end_secs)))
self.sec_range = (start_secs, end_secs)
if anno.properties.get("frameType"):
self.frametype = str(anno.properties.get("frameType"))
@@ -107,24 +112,43 @@ def add_timeframe(self, anno, mmif):
self.frametype = str(anno.properties.get("label"))
def add_timepoint(self, anno, mmif, skip_if_view_has_frames=True):
- parent = mmif.get_view_by_id(anno.parent)
- other_annotations = [k for k in parent.metadata.contains.keys() if k != anno.id]
- # If there are TimeFrames in the same view, they most likely represent
- # condensed information about representative frames (e.g. SWT). In this
- # case, only render the TimeFrames and ignore the TimePoints.
- if any([anno == AnnotationTypes.TimeFrame for anno in other_annotations]) and skip_if_view_has_frames:
- return
- self.frame_num = convert_timepoint(mmif, anno, "frames")
- self.secs = convert_timepoint(mmif, anno, "seconds")
- self.timestamp = str(datetime.timedelta(seconds=self.secs))
- if anno.properties.get("label"):
- self.frametype = anno.properties.get("label")
+ parent = mmif.get_view_by_id(anno.parent)
+ other_annotations = [
+ k for k in parent.metadata.contains.keys() if k != anno.id]
+ # If there are TimeFrames in the same view, they most likely represent
+ # condensed information about representative frames (e.g. SWT). In this
+ # case, only render the TimeFrames and ignore the TimePoints.
+ if any([anno.shortname == "TimeFrame" for anno in other_annotations]) and skip_if_view_has_frames:
+ return
+ self.frame_num = convert_timepoint(mmif, anno, "frames")
+ self.secs = convert_timepoint(mmif, anno, "seconds")
+ self.timestamp = str(datetime.timedelta(seconds=self.secs))
+ if anno.properties.get("label"):
+ self.frametype = anno.properties.get("label")
def add_text_document(self, anno):
- t = anno.properties.get("text_value") or anno.properties.get("text").value
+ t = anno.properties.get(
+ "text_value") or anno.properties.get("text").value
if t:
text_val = re.sub(r'([\\\/\|\"\'])', r'\1 ', t)
- self.text = self.text + [text_val] if text_val not in self.text else self.text
+ self.text = self.text + \
+ [text_val] if text_val not in self.text else self.text
+
+
+def prepare_ocr(mmif, view, viz_id):
+ """
+ Prepares list of frames that will be passed back and forth between server
+ and client, and renders the first page of the OCR.
+ """
+ ocr_frames = get_ocr_frames(view, mmif)
+
+ # Generate pages (necessary to reduce IO cost) and render
+ frames_list = [(k, vars(v)) for k, v in ocr_frames.items()]
+ frames_list = find_duplicates(frames_list)
+ frames_pages = paginate(frames_list)
+ # Save page list as temp file
+ save_json(frames_pages, view.id, viz_id)
+
def get_ocr_frames(view, mmif):
@@ -149,7 +173,7 @@ def get_ocr_frames(view, mmif):
frames[i].update(target, mmif)
else:
frames[i] = frame
-
+
else:
for annotation in view.get_annotations():
frame = OCRFrame(annotation, mmif)
@@ -160,7 +184,6 @@ def get_ocr_frames(view, mmif):
frames[i].update(annotation, mmif)
else:
frames[i] = frame
- print(frames)
return frames
@@ -183,45 +206,9 @@ def paginate(frames_list):
return {i: page for (i, page) in enumerate(pages)}
-def render_ocr(mmif_id, vid_path, view_id, page_number):
- """
- Iterate through frames and display the contents/alignments.
- """
- # Path for storing temporary images generated by cv2
- cv2_vid = cv2.VideoCapture(vid_path)
- tn_data_fname = cache.get_cache_root() / mmif_id / f"{view_id}-pages.json"
- thumbnail_pages = json.load(open(tn_data_fname))
- page = thumbnail_pages[str(page_number)]
- prev_frame_cap = None
- path = make_image_directory(mmif_id)
- for frame_num, frame in page:
- # If index is range instead of frame...
- if frame.get("range"):
- frame_num = (int(frame["range"][0]) + int(frame["range"][1])) / 2
- cv2_vid.set(1, frame_num)
- _, frame_cap = cv2_vid.read()
- if frame_cap is None:
- raise FileNotFoundError(f"Video file {vid_path} not found!")
-
- # Double check histogram similarity of "repeat" frames -- if they're significantly different, un-mark as repeat
- if prev_frame_cap is not None and frame["repeat"] and not is_duplicate_image(prev_frame_cap, frame_cap,
- cv2_vid):
- frame["repeat"] = False
- with tempfile.NamedTemporaryFile(dir=str(path), suffix=".jpg", delete=False) as tf:
- cv2.imwrite(tf.name, frame_cap)
- # "id" is just the name of the temp image file
- frame["id"] = pathlib.Path(tf.name).name
- prev_frame_cap = frame_cap
-
- tn_page_html = render_template(
- 'ocr.html', vid_path=vid_path, view_id=view_id, page=page,
- n_pages=len(thumbnail_pages), page_number=str(page_number), mmif_id=mmif_id)
- return tn_page_html
-
-
-def make_image_directory(mmif_id):
+def make_image_directory(mmif_id, view_id):
# Make path for temp OCR image files or clear image files if it exists
- path = cache.get_cache_root() / mmif_id / "img"
+ path = cache.get_cache_root() / mmif_id / "img" / view_id
if os.path.exists(path):
shutil.rmtree(path)
os.makedirs(path)
@@ -266,10 +253,14 @@ def is_duplicate_image(prev_frame, frame, cv2_vid):
img2_hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
# Calculate the histogram and normalize it
- hist_img1 = cv2.calcHist([img1_hsv], [0, 1], None, [180, 256], [0, 180, 0, 256])
- cv2.normalize(hist_img1, hist_img1, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX);
- hist_img2 = cv2.calcHist([img2_hsv], [0, 1], None, [180, 256], [0, 180, 0, 256])
- cv2.normalize(hist_img2, hist_img2, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX);
+ hist_img1 = cv2.calcHist([img1_hsv], [0, 1], None, [
+ 180, 256], [0, 180, 0, 256])
+ cv2.normalize(hist_img1, hist_img1, alpha=0,
+ beta=1, norm_type=cv2.NORM_MINMAX)
+ hist_img2 = cv2.calcHist([img2_hsv], [0, 1], None, [
+ 180, 256], [0, 180, 0, 256])
+ cv2.normalize(hist_img2, hist_img2, alpha=0,
+ beta=1, norm_type=cv2.NORM_MINMAX)
# Find the metric value
metric_val = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_CHISQR)
@@ -289,29 +280,6 @@ def round_boxes(boxes):
return rounded_boxes
-def get_ocr_views(mmif):
- """Returns all CV views, which contain timeframes or bounding boxes"""
- views = []
- required_types = ["TimeFrame", "BoundingBox", "TimePoint"]
- for view in mmif.views:
- for anno_type, anno in view.metadata.contains.items():
- # Annotation belongs to a CV view if it is a TimeFrame/BB and it refers to a VideoDocument
- # if anno.get("document") is None:
- # continue
- # if anno_type.shortname in required_types and mmif.get_document_by_id(
- # anno["document"]).at_type.shortname == "VideoDocument":
- # views.append(view)
- # continue
- if anno_type.shortname in required_types:
- views.append(view)
- break
- # TODO: Couldn't find a simple way to show if an alignment view is a CV/Frames-type view
- elif "parseq" in view.metadata.app:
- views.append(view)
- break
- return views
-
-
def save_json(data, view_id, mmif_id):
path = cache.get_cache_root() / mmif_id / f"{view_id}-pages.json"
with open(path, 'w') as f:
diff --git a/render.py b/render.py
new file mode 100644
index 0000000..dc4b758
--- /dev/null
+++ b/render.py
@@ -0,0 +1,321 @@
+import os
+import pathlib
+from io import StringIO
+from collections import Counter
+from flask import render_template, current_app
+import re
+
+from mmif import DocumentTypes
+from lapps.discriminators import Uri
+import displacy
+import traceback
+
+from utils import get_status, get_properties, get_abstract_view_type, url2posix, get_vtt_file
+from ocr import prepare_ocr, make_image_directory, is_duplicate_image
+import cv2
+import json
+import tempfile
+
+import cache
+
+"""
+Methods to render MMIF documents and their annotations in various formats.
+"""
+
+# -- Render methods --
+
+
+def render_documents(mmif, viz_id):
+ """
+ Returns HTML Tab representation of all documents in the MMIF object.
+ """
+ tabs = []
+ for document in mmif.documents:
+ if document.at_type == DocumentTypes.TextDocument:
+ tabs.append(TextTab(document, viz_id))
+ elif document.at_type == DocumentTypes.ImageDocument:
+ tabs.append(ImageTab(document, viz_id))
+ elif document.at_type == DocumentTypes.AudioDocument:
+ tabs.append(AudioTab(document, viz_id))
+ elif document.at_type == DocumentTypes.VideoDocument:
+ tabs.append(VideoTab(document, mmif, viz_id))
+
+ return tabs
+
+
+def render_annotations(mmif, viz_id):
+ """
+ Returns HTML Tab representation of all annotations in the MMIF object.
+ """
+ tabs = []
+ # These tabs should always be present
+ tabs.append(InfoTab(mmif))
+ tabs.append(AnnotationTableTab(mmif))
+ tabs.append(JSTreeTab(mmif))
+ # These tabs are optional
+ for view in mmif.views:
+ abstract_view_type = get_abstract_view_type(view, mmif)
+ if abstract_view_type == "NER":
+ tabs.append(NERTab(mmif, view))
+ elif abstract_view_type == "ASR":
+ tabs.append(VTTTab(mmif, view, viz_id))
+ elif abstract_view_type == "OCR":
+ tabs.append(OCRTab(mmif, view, viz_id))
+
+ return tabs
+
+
+# -- Base Tab Class --
+
+class DocumentTab():
+ def __init__(self, document, viz_id):
+ self.id = document.id
+ self.tab_name = document.at_type.shortname
+ self.viz_id = viz_id
+
+ try:
+ # Add symbolic link to document to static folder, so it can be accessed
+ # by the browser.
+ self.doc_path = document.location_path()
+ self.doc_symlink_path = pathlib.Path(
+ current_app.static_folder) / cache._CACHE_DIR_SUFFIX / viz_id / (f"{document.id}.{self.doc_path.split('.')[-1]}")
+ os.symlink(self.doc_path, self.doc_symlink_path)
+ self.doc_symlink_rel_path = '/' + \
+ self.doc_symlink_path.relative_to(
+ current_app.static_folder).as_posix()
+
+ self.html = self.render()
+
+ except Exception as e:
+ self.html = f"Error rendering document:
{traceback.format_exc()}"
+
+ def __str__(self):
+ return f"Tab: {self.tab_name} ({self.id})"
+
+
+class AnnotationTab():
+ def __init__(self, mmif, view=None):
+ self.mmif = mmif
+ # Some AnnotationTab sub-classes don't refer to a specific view, and so
+ # they specify their own ids and tab names. For ones that do refer to
+ # a specific view, we set the ids/tab names based on view properties.
+ if view:
+ self.view = view
+ # Workaround to deal with the fact that some apps have a version number
+ # in the URL
+ app_url = view.metadata.app if re.search(
+ r"\/v\d+\.?\d?$", view.metadata.app) else view.metadata.app + "/v1"
+ app_shortname = app_url.split("/")[-2]
+
+ self.id = view.id
+ self.tab_name = f"{app_shortname}-{view.id}"
+ try:
+ self.html = self.render()
+ except Exception as e:
+ self.html = f"Error rendering view: {traceback.format_exc()}"
+
+
+# -- Document Classes --
+
+class TextTab(DocumentTab):
+ def __init__(self, document, viz_id):
+ super().__init__(document, viz_id)
+
+ def render(self):
+ with open(self.doc_path) as t_file:
+ content = t_file.read().replace("\n", "")
+ for document in mmif.documents:
+ at_type = document.at_type.shortname
+ location = document.location
+ s.write("%s %s\n" % (at_type, location))
+ s.write('\n')
+ for view in mmif.views:
+ app = view.metadata.app
+ status = get_status(view)
+ s.write('%s %s %s %d\n' %
+ (view.id, app, status, len(view.annotations)))
+ if len(view.annotations) > 0:
+ s.write('\n')
+ types = Counter([a.at_type.shortname
+ for a in view.annotations])
+ for attype, count in types.items():
+ s.write(' %4d %s\n' % (count, attype))
+ s.write('\n')
+ s.write("")
+ return s.getvalue()
+
+
+class AnnotationTableTab(AnnotationTab):
+ def __init__(self, mmif):
+ self.id = "annotations"
+ self.tab_name = "Annotations"
+ super().__init__(mmif)
+
+ def render(self):
+ mmif = self.mmif
+ s = StringIO('Howdy')
+ for view in mmif.views:
+ status = get_status(view)
+ s.write('%s %s %s %d annotations
\n' + % (view.id, view.metadata.app, status, len(view.annotations))) + s.write("\n") + s.write("\n") + return s.getvalue() + + +class JSTreeTab(AnnotationTab): + def __init__(self, mmif): + self.id = "tree" + self.tab_name = "Tree" + super().__init__(mmif) + + def render(self): + mmif = self.mmif + return render_template('interactive.html', mmif=mmif, aligned_views=[]) + + +class NERTab(AnnotationTab): + def __init__(self, mmif, view): + super().__init__(mmif, view) + + def render(self): + metadata = self.view.metadata.contains.get(Uri.NE) + ner_document = metadata.get('document') + return displacy.visualize_ner(self.mmif, self.view, ner_document, current_app.root_path) + + +class VTTTab(AnnotationTab): + def __init__(self, mmif, view, viz_id): + self.viz_id = viz_id + super().__init__(mmif, view) + + def render(self): + vtt_filename = get_vtt_file(self.view, self.viz_id) + with open(vtt_filename) as vtt_file: + vtt_content = vtt_file.read() + return f"\n") + def limit_len(str): return str[:500] + \ + " . . . }" if len(str) > 500 else str + for annotation in view.annotations: + s.write('
\n") + s.write("\n') + s.write(' \n') + s.write("%s \n' % annotation.id) + s.write('%s \n' % annotation.at_type.shortname) + s.write('%s \n' % + limit_len(get_properties(annotation))) + s.write('
{vtt_content}"
+
+
+class OCRTab(AnnotationTab):
+ def __init__(self, mmif, view, viz_id):
+ self.viz_id = viz_id
+ self.vid_path = mmif.get_documents_by_type(DocumentTypes.VideoDocument)[
+ 0].location_path()
+
+ super().__init__(mmif, view)
+
+ def render(self):
+ return render_template("pre-ocr.html", view_id=self.view.id, tabname=self.tab_name, mmif_id=self.viz_id)
+ # prepare_ocr(self.mmif, self.view, self.viz_id)
+ # return render_ocr_page(self.viz_id, self.vid_path, self.view.id, 0)
+
+
+def render_ocr_page(mmif_id, vid_path, view_id, page_number):
+ """
+ Renders a single OCR page by iterating through frames and displaying the
+ contents/alignments. Note: this needs to be a separate function (not a method
+ in OCRTab) because it is called by the server when the page is changed.
+ """
+ # Path for storing temporary images generated by cv2
+ cv2_vid = cv2.VideoCapture(vid_path)
+ tn_data_fname = cache.get_cache_root() / mmif_id / f"{view_id}-pages.json"
+ thumbnail_pages = json.load(open(tn_data_fname))
+ page = thumbnail_pages[str(page_number)]
+ prev_frame_cap = None
+ path = make_image_directory(mmif_id, view_id)
+ for frame_num, frame in page:
+ # If index is range instead of frame...
+ if frame.get("range"):
+ frame_num = (int(frame["range"][0]) + int(frame["range"][1])) / 2
+ cv2_vid.set(1, frame_num)
+ _, frame_cap = cv2_vid.read()
+ if frame_cap is None:
+ raise FileNotFoundError(f"Video file {vid_path} not found!")
+
+ # Double check histogram similarity of "repeat" frames -- if they're significantly different, un-mark as repeat
+ if prev_frame_cap is not None and frame["repeat"] and not is_duplicate_image(prev_frame_cap, frame_cap,
+ cv2_vid):
+ frame["repeat"] = False
+ with tempfile.NamedTemporaryFile(dir=str(path), suffix=".jpg", delete=False) as tf:
+ cv2.imwrite(tf.name, frame_cap)
+ # "id" is just the name of the temp image file
+ frame["id"] = pathlib.Path(tf.name).name
+ prev_frame_cap = frame_cap
+
+ tn_page_html = render_template(
+ 'ocr.html', vid_path=vid_path, view_id=view_id, page=page,
+ n_pages=len(thumbnail_pages), page_number=str(page_number), mmif_id=mmif_id)
+ return tn_page_html
diff --git a/templates/ocr.html b/templates/ocr.html
index 7daea2b..9bf370f 100644
--- a/templates/ocr.html
+++ b/templates/ocr.html
@@ -1,7 +1,7 @@
{{ docs[0][2] }}
- {{ docs[0][3] | safe }} + + {{ docs[0]['html'] | safe }}{{ medium[2] }}
- {{ medium[3] | safe }} + + {{ medium['html'] | safe }}" + create_info(mmif) + "")) - app.logger.debug(f"Prepared INFO Tab: {tabs[-1][0]}") - # tabs.append(("MMIF", "
" + mmif.serialize(pretty=True) + "")) - # app.logger.debug(f"Prepared RAW Tab: {tabs[-1][0]}") - tabs.append(("Annotations", create_annotation_tables(mmif))) - app.logger.debug(f"Prepared SUMMARY Tab: {tabs[-1][0]}") - tabs.append(("Tree", render_interactive_mmif(mmif))) - app.logger.debug(f"Prepared JSTREE Tab: {tabs[-1][0]}") - # TODO: since this uses the same tab-name this will only show the same - # stuff; it does a loop but for now we assume there is just one file with - # alignments (generated by Kaldi) - for fa_view in get_alignment_views(mmif): - vtt_file = asr_alignments_to_vtt(fa_view, viz_id) - tabs.append(("WebVTT", '
' + open(vtt_file).read() + '')) - app.logger.debug(f"Prepared a VTT Tab: {tabs[-1][0]}") - ner_views = get_ner_views(mmif) - use_id = True if len(ner_views) > 1 else False - for ner_view in ner_views: - if not ner_view.annotations: - continue - visualization = create_ner_visualization(mmif, ner_view) - tabname = "Entities-%s" % ner_view.id if use_id else "Entities" - tabs.append((tabname, visualization)) - app.logger.debug(f"Prepared a displaCy Tab: {tabs[-1][0]}") - # TODO: somewhat hackish - ocr_views = get_ocr_views(mmif) - use_id = True if len(ocr_views) > 1 else False - for ocr_view in ocr_views: - if not ocr_view.annotations: - continue - tabname = "Thumbnails-%s" % ocr_view.id - visualization = render_template("pre-ocr.html", view_id=ocr_view.id, tabname=tabname, mmif_id=viz_id) - tabs.append((tabname, visualization)) - app.logger.debug(f"Prepared a Thumbnails Tab: {tabs[-1][0]}") - return tabs - - -def create_info(mmif): - s = StringIO('Howdy') - for document in mmif.documents: - at_type = document.at_type.shortname - location = document.location - s.write("%s %s\n" % (at_type, location)) - s.write('\n') - for view in mmif.views: - app = view.metadata.app - status = get_status(view) - s.write('%s %s %s %d\n' % (view.id, app, status, len(view.annotations))) - if len(view.annotations) > 0: - s.write('\n') - types = Counter([a.at_type.shortname - for a in view.annotations]) - for attype, count in types.items(): - s.write(' %4d %s\n' % (count, attype)) - s.write('\n') - return s.getvalue() - - -def create_annotation_tables(mmif): - s = StringIO('Howdy') - for view in mmif.views: - status = get_status(view) - s.write('
%s %s %s %d annotations
\n' - % (view.id, view.metadata.app, status, len(view.annotations))) - s.write("\n") - s.write("\n") - return s.getvalue() - - -def get_document_ids(view, annotation_type): - metadata = view.metadata.contains.get(annotation_type) - ids = set([metadata['document']]) if 'document' in metadata else set() - for annotation in view.annotations: - if annotation.at_type.shortname == str(annotation_type): - try: - ids.add(annotation.get("document")) - except KeyError: - pass - return list(ids) - - -def get_alignment_views(mmif): - """Return alignment views which have at least TextDocument, Token, TimeFrame and - Alignment annotations.""" - views = [] - needed_types = set(['TextDocument', 'Token', 'TimeFrame', 'Alignment']) - for view in mmif.views: - annotation_types = view.metadata.contains.keys() - annotation_types = [at.shortname for at in annotation_types] - if needed_types.issubset(annotation_types): - views.append(view) - return views - - -# Render documents as HTML ------------ - -def html_video(viz_id, vpath, vtt_srcview=None): - vpath = url2posix(vpath) - html = StringIO() - html.write('\n") - return html.getvalue() - - -def html_text(tpath): - """Return the content of the text document, but with some HTML tags added.""" - if not os.path.isfile(tpath): - raise FileNotFoundError(f"File not found: {tpath}") - with open(tpath) as t_file: - content = t_file.read().replace("\n", "\n") - limit_len = lambda str: str[:500] + " . . . }" if len(str) > 500 else str - for annotation in view.annotations: - s.write('
\n") - s.write("\n') - s.write(' \n') - s.write("%s \n' % annotation.id) - s.write('%s \n' % annotation.at_type.shortname) - s.write('%s \n' % limit_len(get_properties(annotation))) - s.write('