Comments (10)
Hi @Marouene-Oueslati,
Try adding the line
imdb.extract_and_save_images(extract_frame_type='annotated')
after line 56 (imdb = PIE(data_path=pie_path)
) in train_test.py
. This should extract the images. You will need 1.1 Tb of space to hold all images from the dataset with annotations. If you are not planning to use training and validation data, you can modify the function pie_data.py:extract_and_save_images()
to extract images only from set03
(test data).
from piepredict.
Hi @ykotseruba ,
i'm trying to extract only from set01
as test data, the following issues appear on the command line when python train_test.py 1
runs:
Generating train features crop_type=context crop_mode=pad_resize
save_path=./data/data\pie\train\features_context_pad_resize\vgg16_none,
[--------------------] 0.00% Traceback (most recent call last):
File "train_test.py", line 178, in <module>
main(train_test=train_test)
File "train_test.py", line 171, in main
intent_model_path = train_intent(train_test=train_test)
File "train_test.py", line 149, in train_intent
data_opts=data_opts)
File "C:\Users\Jar\object_detection\PIEPredict\pie_intent.py", line 550, in train
data_subset = 'train'))
File "C:\Users\Jar\object_detection\PIEPredict\pie_intent.py", line 242, in load_images_and_process
set_id = imp.split('/')[-3]
IndexError: list index out of range
could you please help me with this? Thanks!
from piepredict.
Hi @RoteOrange,
The error is likely caused by path formatting: paths in Windows are separated with \
and in Linux with /
. To fix this you can try changing imp.split('/')
to imp.split('\')
.
In general, it is better to run the code on Ubuntu where it was tested. Unfortunately, I don't have access to a Windows machine, so if there are other Windows-specific issues, I won't be able to help.
from piepredict.
Hi @Marouene-Oueslati,
Try adding the lineimdb.extract_and_save_images(extract_frame_type='annotated')
after line 56 (
imdb = PIE(data_path=pie_path)
) intrain_test.py
. This should extract the images. You will need 1.1 Tb of space to hold all images from the dataset with annotations. If you are not planning to use training and validation data, you can modify the functionpie_data.py:extract_and_save_images()
to extract images only fromset03
(test data).
Hi @ykotseruba ,
Thanks a lot for your answer. I have configured the right set of video clips.
That being said, do we need to use sh function "split_clips_to_frames.sh" to extract the needed images for the desired Set-XX beforehand as I still have the same issue.
Furthermore, I am wondering if the model could tested for any type of raw video without being linked to annotated images (I mean in Real Time application)
Thanks a lot for your valuable feedback
from piepredict.
Hi @Marouene-Oueslati,
split_clips_to_frames.sh
extracts all frames, whereas the line of code I suggested extracts only the annotated frames (1/3 of all frames). If nothing else works, you may try using the bash script but it will require 3 times more memory.
To answer your second question, the model needs bounding boxes for pedestrians to work. In order to apply it to videos that are not part of dataset you will either have to annotate them manually or implement object detection/tracking. For evaluation you will also need other ground truth: intention scores, vehicle information, etc.
from piepredict.
Hi @ykotseruba ,
your feedback is highly appreciated. Indeed I do understand your feedback as the code as well as the paper was part of a deep research topic with in-house validation.
The line of code till now seems not working, perhaps I will try to extract it accordingly using the split_clips_to_frames.sh.
Just wondering if you have any idea regarding the use of yolo / ss to extract the bounding boxes & reshaping the BB convention.
My last question concerns the possibility of only training the intention model without linking it to trajectory prediction nor vehicle speed; if so, could you please share some rules or approach.
My apologies for the long message.
from piepredict.
Hi @Marouene-Oueslati,
Could you please be more specific? What error do you get when running the extract_and_save_images()
?
Regarding two other questions, I didn't run object detectors on PIE so I can't help you with this. If you need annotations in YOLO or any other format you can use CVAT to convert them.
It is possible to train only the intent model. train_intent()' function has code for both training and testing. Try commenting out
train_predict()in
train_test.py:173`, that should probably do it.
from piepredict.
Hello @ykotseruba; this is a screen shot of the error that I am getting when running the "python train_test.py 2". FYI, as you can see, I am only testing for video 0002 from the set01 and have added imdb.extract_and_save_images(extract_frame_type='annotated')
from piepredict.
Hi @Marouene-Oueslati,
Based on the error, the images aren't extracted. Do you get any output or errors from extract_and_save_images
?
from piepredict.
Closing due to inactivity.
from piepredict.
Related Issues (17)
- PIEPredict vs. SF-GRU HOT 2
- how to make get vehicle annotations HOT 3
- Trajectory Prediction HOT 2
- memory error HOT 1
- Comment trouver 79% de précision
- Bounding box normalization HOT 8
- De-normalizing bounding box coordinates HOT 2
- Intention Data in Test Stage
- config.json
- Allocation of 9707550720 exceeds 10% of free system memory. Error
- pie_path HOT 2
- dataset HOT 6
- How to generate annotations_attributes for custom dataset HOT 2
- how to use the model HOT 4
- trajectory HOT 1
- PIE HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from piepredict.