ddp/photoassistant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

quick fix/hack for unmatched faces

Browse Source
This commit is contained in:
Damien De Paoli
2022-07-19 21:23:32 +10:00
parent fd79ee2cf4
commit 8646b3b8c2
3 changed files with 27 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
TODO
View File

@@ -1,34 +1,31 @@
## GENERAL
* on viewer:
- allow face to be used to:
[DONE] - create person
[DONE] - add to existing person
--> still need to consider whether we trigger an AI search immediately (for these 2 options)
[DONE] - create person
[DONE] - add to existing person
--> still need to consider whether we trigger an AI search immediately (for these 2 options)
[DONE] - ignore/not a face/too young
[DONE] - redraw 'ignore's as a greyed out box?
[DONE] - menu should only allow override IF we have put override on...
all NMO's need to handle delete data and rebuild / allow recreation of content form FS (not just test, it causes a bug now / db constraint violation)
--> need to test the 'override' when we re-ai-match (AFTER re-build from FS)
all NMO's need to handle delete data and rebuild / allow recreation of content form FS (not just test, it causes a bug now / db constraint violation)
--> need to test the 'override' when we re-ai-match (AFTER re-build from FS)
* capture pa_job_manager logs in prod
- redir'd to /var/log/pa* but NO content, seems wrong
* capture pa_job_manager logs in prod
- redir'd to /var/log/pa* but NO content, seems wrong
* should I change the rotation code to use that jpeg util to reduce/remove compression loss?
* should I change the rotation code to use that jpeg util to reduce/remove compression loss?
* should allow context menu from View thumbs (particularly useful on search) to show other files around this one by date (maybe that folder or something?)
* re-think unmatched faces...
(the view needs a orig_url and pa_user_state to support viewer for current)
* could get better AI optim, by keeping track of just new files since scan (even if we did this from the DB),
then we could just feed those eid's explicitly into a 'run_ai_on_new_files' :) -- maybe particularly
if count('new files') < say 1000 do eids, otherwise do path AND no new refimgs
* does search of matching dirname give all entries of subdirs of subdirs, etc. (think not) -- maybe a TODO?
* video player cannot handle non mp4 formats... do I care? (could just offer a download link and hope the client deals with it)
--> OR? https://jsmpeg.com/
--> OR? convert all videos to mp4/webm
* video player cannot handle non mp4 formats... do I care? (could just offer a download link and hope the client deals with it)
--> OR? https://jsmpeg.com/
--> OR? convert all videos to mp4/webm
* delete folder
@@ -50,7 +47,7 @@
- searched for text overlaps buttons above and below
- < 10 files > is subsequently not centered
- the folder/bin icons might be best below search then? (and on same line as XS/S, etc.)
* on phone login page, job* pages, etc. are all squished
* on phone login page, job* pages, etc. are all squished
* when search, have a way to hide deleted files
-> not sure where to put this on GUI, its so busy...
@@ -73,13 +70,13 @@
https://www.pyimagesearch.com/2018/07/09/face-clustering-with-python/
* fix up logging in general
ProcessFileForJob --> really need to better handle log levels and counting
ProcessFileForJob --> really need to better handle log levels and counting
* support animated gifs in html5 canvas
* think about security - in job_mgr anywhere I can os.replace/remove NEED to protect, etc
## DB
## DB
* Dir can have date in the DB, so we can do Oldest/Newest dirs in Folder view
### BACKEND
@@ -118,8 +115,8 @@
### SORTER
* exif processing?
* location stuff - test a new photo from my camera out
* exif processing?
* location stuff - test a new photo from my camera out
-- image is in dir, need to look at exifread output
### FUTURE:

11
ai.py
View File

@@ -87,14 +87,15 @@ def unmatched_faces():
faces=Face.query.join(FaceFileLink).join(FaceRefimgLink, isouter=True).filter(FaceRefimgLink.refimg_id==None).order_by(Face.h.desc()).limit(10).all()
imgs={}
for face in faces:
tmp_locn=json.loads(face.locn)
face.tmp_locn=json.loads(face.locn)
f = Entry.query.join(File).join(FaceFileLink).filter(FaceFileLink.face_id==face.id).first()
face.file_eid=f.id
face.url=f.FullPathOnFS()
x=tmp_locn[3]*0.95
y=tmp_locn[0]*0.95
x2=tmp_locn[1]*1.05
y2=tmp_locn[2]*1.05
face.filename=f.name
x=face.tmp_locn[3]*0.95
y=face.tmp_locn[0]*0.95
x2=face.tmp_locn[1]*1.05
y2=face.tmp_locn[2]*1.05
im = Image.open(f.FullPathOnFS())
region = im.crop((x, y, x2, y2))

12
templates/faces.html
View File

@@ -7,9 +7,7 @@
<div class="row mt-3">
{% for f in faces %}
<div id="F{{f.id}}" class="col col-auto mt-3 pr-1">
<form id="_fm" method="POST" action="/view/{{f.file_eid}}">
<input type="hidden" name="eids" value="{{f.file_eid}},">
<input type="hidden" name="orig_url" value="/files_ip">
<form id="_fm" method="POST" action="/search/{{f.filename}}">
<figure id="fig_{{f.id}}" class="mb-0">
<div style="position:relative">
<canvas id="c_{{f.id}}" height="128"></canvas>
@@ -19,12 +17,12 @@
fig_{{f.id}}=$('#fig_{{f.id}}')
// store this stuff in an javascript Object to use when document is ready event is triggered
var orig_face_{{f.id}}=new Object;
orig_face_{{f.id}}.x = (({{f.locn[1]}}*1.05 - {{f.locn[3]}}*.95) - {{f.w}}) / 2
orig_face_{{f.id}}.y = (({{f.locn[2]}}*1.05 - {{f.locn[0]}}*.95) - {{f.h}}) / 2
orig_face_{{f.id}}.x = (({{f.tmp_locn[1]}}*1.05 - {{f.tmp_locn[3]}}*.95) - {{f.w}}) / 2
orig_face_{{f.id}}.y = (({{f.tmp_locn[2]}}*1.05 - {{f.tmp_locn[0]}}*.95) - {{f.h}}) / 2
orig_face_{{f.id}}.w = {{f.w}}
orig_face_{{f.id}}.h = {{f.h}}
orig_face_{{f.id}}.orig_w = {{f.locn[1]}}*1.05 - {{f.locn[3]}}*.95
orig_face_{{f.id}}.orig_h = {{f.locn[2]}}*1.05 - {{f.locn[0]}}*.95
orig_face_{{f.id}}.orig_w = {{f.tmp_locn[1]}}*1.05 - {{f.tmp_locn[3]}}*.95
orig_face_{{f.id}}.orig_h = {{f.tmp_locn[2]}}*1.05 - {{f.tmp_locn[0]}}*.95
//console.log( orig_face_{{f.id}} )