yassersouri / fast-bird-part-localization Goto Github PK
View Code? Open in Web Editor NEWCode for Fast Bird Part Localization (FGVC 2015)
Home Page: http://yassersouri.github.io/pages/fast-bird-part.html
License: MIT License
Code for Fast Bird Part Localization (FGVC 2015)
Home Page: http://yassersouri.github.io/pages/fast-bird-part.html
License: MIT License
The "body" detector training bug log:
fast_bird_part_localization/geometry_utils.py:59: RuntimeWarning: overflow encountered in ubyte_scalars
cenX = float(self.xmin + self.xmax) / 2
fast_bird_part_localization/geometry_utils.py:60: RuntimeWarning: overflow encountered in ubyte_scalars
cenY = float(self.ymin + self.ymax) / 2
done 0 in 0:00:00.000088
done 100 in 0:07:55.448850
...
done 2400 in 1:50:29.576296
done 2500 in 1:55:50.305564
Traceback (most recent call last):
File "create_rf_model.py", line 65, in
main()
File "create_rf_model.py", line 36, in main
pos, neg = dh.image_point_features(img, part_box, part_name)
File "fast_bird_part_localization/cnn_utils.py", line 108, in image_point_features
return self.features(positive_points), self.features(negative_points)
File "fast_bird_part_localization/cnn_utils.py", line 91, in features
features[i, :] = self.ffeats[point[0], point[1], :]
IndexError: index 341 is out of bounds for axis 0 with size 333
after running "head, head_prob = fbp.detect(img)", I have got these errors:
head, head_prob = fbp.detect(img)
Traceback (most recent call last):
File "", line 1, in
File "fast_bird_part_localization/detector.py", line 23, in detect
pred_box = geometry_utils.find_rect_from_preds(geometry_utils.post_process_preds(preds_img))
File "fast_bird_part_localization/geometry_utils.py", line 205, in post_process_preds
preds = skimage.morphology.closing(preds, skimage.morphology.square(10))
File "/usr/lib/python2.7/dist-packages/skimage/morphology/grey.py", line 215, in closing
dilated = dilation(image, selem)
File "/usr/lib/python2.7/dist-packages/skimage/morphology/grey.py", line 112, in dilation
image = img_as_ubyte(image)
File "/usr/lib/python2.7/dist-packages/skimage/util/dtype.py", line 358, in img_as_ubyte
return convert(image, np.uint8, force_copy)
File "/usr/lib/python2.7/dist-packages/skimage/util/dtype.py", line 99, in convert
raise ValueError("can not convert %s to %s." % (dtypeobj_in, dtypeobj))
ValueError: can not convert int64 to uint8.
flatten makes a copy of the data
Hi,
Nice code! I was wondering which type of split function is used by scikit-learn library, any idea? I am not sure if typically used value/pixel difference split function will work well with hyper-column features, since there is no correlation between feature values.
Thanks!
The 'bbox' detector training bug: if one changes the part name from 'head' to 'bbox' at
https://github.com/yassersouri/fast-bird-part-localization/blob/master/create_rf_model.py#L23
However, because the parameters setting dict has no key-value pair for 'bbox' at:
https://github.com/yassersouri/fast-bird-part-localization/blob/master/fast_bird_part_localization/settings.py#L10
So the training for 'bbox' will fail at:
https://github.com/yassersouri/fast-bird-part-localization/blob/master/fast_bird_part_localization/cnn_utils.py#L103
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.