search

obendidi / Tracking-with-darkflow

Real-time people Multitracker using YOLO v2 and deep_sort with tensorflow
GNU General Public License v3.0
524 stars 174 forks source link

Performance in terms of objects #23

Closed ashnarayan13 closed 6 years ago

ashnarayan13 commented 6 years ago

I wanted to know how many objects can be tracked in each frame as the README doesn't clearly state it.

Also, I wanted to know if this can be extended to track objects of different classes. Like, people, animals simultaneously. (Yolo can do that, but can the tracking keep the IDs?)

Thanks

obendidi commented 6 years ago

I don't think there is a limit to the number of objects tracked in each frame, and the current implementation can track objects of different classes (by giving a list of class names to the trackObj FLAG) but it will not keep the name of the classes while tracking , it will just show the tracking number ID

ashnarayan13 commented 6 years ago

Thanks! Also, what weights do I use, using YOLO weights is giving a tensor flow error. The yolo weights from the official darknet website if yolo weights for the C and CUDA implementation, where can I get the Tensorflow weights and cfg?

FLAGS.model = "darkflow/cfg/yolo.cfg" # tensorflow model

FLAGS.load = "darkflow/bin/yolo.weights" # tensorflow weights

FLAGS.pbLoad = "tiny-yolo-voc-traffic.pb" # tensorflow model FLAGS.metaLoad = "tiny-yolo-voc-traffic.meta" # tensorflow weights

Thanks again!

obendidi commented 6 years ago

you can get YOLO weights from here, the cfg are already there by default

ashnarayan13 commented 6 years ago

Thanks. The weights don't seem to be a problem. Tensorflow is not built for CUDA 9 yet and hence the run.py is not working. Even after setting GPU to 0. I get a tensorflow error

Traceback (most recent call last): File "run.py", line 3, in from darkflow.darkflow.net.build import TFNet File "/home/ashwath/fortiss/Tracking-with-darkflow/darkflow/darkflow/net/build.py", line 1, in import tensorflow as tf File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/init.py", line 24, in from tensorflow.python import File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/python/init.py", line 49, in from tensorflow.python import pywrap_tensorflow File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 72, in raise ImportError(msg) ImportError: Traceback (most recent call last): File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in from tensorflow.python.pywrap_tensorflow_internal import File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 28, in _pywrap_tensorflow_internal = swig_import_helper() File "/home/ashwath/.local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py", line 24, in swig_import_helper _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description) ImportError: libcublas.so.8.0: cannot open shared object file: No such file or directory

Failed to load the native TensorFlow runtime.

obendidi commented 6 years ago

the error is because you are using tensorflow-gpu , and yes tensorflow does not support cuda-9 yet , you can install cuda-8 simply using apt-get install cuda-8.0

ashnarayan13 commented 6 years ago

Got it working on CPU instead. Thank you so much for the quick response!