eg4000
commented
4 years ago For test mode add "convert=True" (see file predict.py) model = models.load_model(model_path, backbone_name='resnet50',convert=True)
Closed RafayAK closed 4 years ago
eg4000
commented
4 years ago For test mode add "convert=True" (see file predict.py) model = models.load_model(model_path, backbone_name='resnet50',convert=True)
RafayAK
commented
4 years ago LOL! That was it!
RafayAK
commented
4 years ago ...+ pands 1.0.xx and above don't work since .ix has been consistently used in EmMerger.py. Fixed that by downgrading to 0.25.xx
Thanks a lot, Eran! I'll see If I can convert this model to PyTorch for some experiments.
CodeJjang
commented
4 years ago @RafayAK Didn't you receive the following error while using the pretrained model?
cv2.error: OpenCV(4.1.0) /io/opencv/modules/imgproc/src/shapedescr.cpp:743: error: (-215:Assertion failed) npoints >= 0 && (depth == CV_32F || depth == CV_32S) in function 'pointSetBoundingRect'I saw the fix in #30 but it was related to the remove_redundant function and here it occurs in find_new_candidates, on the following line:
contour_bounding_rect = cv2.boundingRect(contour)Hi @CodeJjang. Did you try the following change for opencv4?
for contour_i, contour in enumerate(contours[0]):
contour_bounding_rect = cv2.boundingRect(contour)@eg4000 I applied that now, the cnts.append(contours[0][0]) fix and another small fix for an old python 2 iteration.
However the EM doesn't seem to converge, I get this output:
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
agglomerative_init Timeout - using fallback
1612 78 k<=Params.min_k or EM failed
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
EM did not converge- using fallback
...Using your pretrained weights.
eg4000
commented
4 years ago It might use fallback in rare cases.
RafayAK
commented
4 years ago @CodeJjang Sorry for the late reply, unfortunately, I didn't face the same problems as you. I'll see If I can attach a sample code showing the way I tested the author's model. I'll also attach my environment setup perhaps that may come in handy.
RafayAK
commented
4 years ago @CodeJjang
substitute ENVIRONMNET_NAME with the name you want to give to this virtual environment
conda create -n ENVIRONMNET_NAME python=3.6.5 opencv=3.4
conda activate ENVIRONMNET_NAME
conda install -c anaconda tensorflow=1.15
conda install -c conda-forge keras=2.2.4
conda install -c conda-forge scipy=1.41
conda install -c conda-forge pandas=0.25.1
conda install -c conda-forge pillow
conda install -c conda-forge tqdm
conda install -c conda-forge matplotlib
pip install keras-resnet
conda install -c conda-forge jupyterlab (optional)
RafayAK
commented
4 years ago @CodeJjang https://gist.github.com/RafayAK/98988d3bbb2d1160b11a0b1d5e719da5 example of inference
gmt710
commented
4 years ago Hi Eran, I tried to test out the pre-trained model that you graciously provided in #9, but I can't seem to figure out why it does not work? Following is the way I'm trying to make it work:
# import keras_retinanet from object_detector_retinanet.keras_retinanet import models from object_detector_retinanet.keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image from object_detector_retinanet.keras_retinanet.utils.visualization import draw_box, draw_caption from object_detector_retinanet.keras_retinanet.utils.colors import label_color # import for EM Merger and viz from object_detector_retinanet.keras_retinanet.utils import EmMerger from object_detector_retinanet.utils import create_folder, root_dir # import miscellaneous modules import matplotlib.pyplot as plt import cv2 import os import numpy as np import time # set tf backend to allow memory to grow, instead of claiming everything import tensorflow as tf model_path = 'iou_resnet50_csv_06.h5' model = models.load_model(model_path, backbone_name='resnet50') image = read_image_bgr("test.jpg") # for filtering predictions based on score (objectness/confidence) threshold = 0.3 # copy to draw on draw = image.copy() draw = cv2.cvtColor(draw, cv2.COLOR_BGR2RGB) # preprocess image for network image = preprocess_image(image) image, scale = resize_image(image) # Run inference boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0)) ... ...... this is where the inference fails:
boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0)) ValueError: not enough values to unpack (expected 4, got 3)I've been looking at previously closed issues in solving this #21, #51, and #30, sadly to no avail. Any help will be much appreciated
PS: Following are my dependencies setup:
# Name Version Build Channel _libgcc_mutex 0.1 main _tflow_select 2.3.0 mkl anaconda absl-py 0.9.0 py36_0 anaconda astor 0.8.0 py36_0 anaconda attrs 19.3.0 py_0 conda-forge backcall 0.1.0 py_0 conda-forge blas 1.0 mkl bleach 3.1.4 pyh9f0ad1d_0 conda-forge bzip2 1.0.8 h7b6447c_0 c-ares 1.15.0 h7b6447c_1001 anaconda ca-certificates 2020.4.5.1 hecc5488_0 conda-forge cairo 1.14.12 h8948797_3 certifi 2020.4.5.1 py36h9f0ad1d_0 conda-forge cycler 0.10.0 py_2 conda-forge dbus 1.13.6 he372182_0 conda-forge decorator 4.4.2 py_0 conda-forge defusedxml 0.6.0 py_0 conda-forge entrypoints 0.3 py36h9f0ad1d_1001 conda-forge expat 2.2.9 he1b5a44_2 conda-forge ffmpeg 4.0 hcdf2ecd_0 fontconfig 2.13.0 h9420a91_0 freeglut 3.0.0 hf484d3e_5 freetype 2.9.1 h8a8886c_1 gast 0.2.2 py36_0 anaconda glib 2.63.1 h5a9c865_0 google-pasta 0.2.0 py_0 anaconda graphite2 1.3.13 h23475e2_0 grpcio 1.14.1 py36h9ba97e2_0 anaconda gst-plugins-base 1.14.5 h0935bb2_2 conda-forge gstreamer 1.14.5 h36ae1b5_2 conda-forge h5py 2.8.0 py36h989c5e5_3 anaconda harfbuzz 1.8.8 hffaf4a1_0 hdf5 1.10.2 hba1933b_1 icu 58.2 h9c2bf20_1 importlib-metadata 1.6.0 py36h9f0ad1d_0 conda-forge importlib_metadata 1.6.0 0 conda-forge intel-openmp 2020.0 166 ipykernel 5.2.1 py36h95af2a2_0 conda-forge ipython 7.13.0 py36h9f0ad1d_2 conda-forge ipython_genutils 0.2.0 py_1 conda-forge jasper 2.0.14 h07fcdf6_1 jedi 0.17.0 py36h9f0ad1d_0 conda-forge jinja2 2.11.2 pyh9f0ad1d_0 conda-forge jpeg 9b h024ee3a_2 json5 0.9.0 py_0 conda-forge jsonschema 3.2.0 py36h9f0ad1d_1 conda-forge jupyter_client 6.1.3 py_0 conda-forge jupyter_core 4.6.3 py36h9f0ad1d_1 conda-forge jupyterlab 2.1.0 py_0 conda-forge jupyterlab_server 1.1.1 py_0 conda-forge keras 2.2.4 py36_1 conda-forge keras-applications 1.0.8 py_0 anaconda keras-preprocessing 1.1.0 py_1 anaconda keras-resnet 0.2.0 pypi_0 pypi kiwisolver 1.2.0 py36hdb11119_0 conda-forge libblas 3.8.0 15_mkl conda-forge libcblas 3.8.0 15_mkl conda-forge libedit 3.1.20181209 hc058e9b_0 libffi 3.2.1 hd88cf55_4 libgcc-ng 9.1.0 hdf63c60_0 libgfortran-ng 7.3.0 hdf63c60_0 libglu 9.0.0 hf484d3e_1 libgpuarray 0.7.6 h14c3975_1003 conda-forge liblapack 3.8.0 15_mkl conda-forge libopencv 3.4.2 hb342d67_1 libopus 1.3.1 h7b6447c_0 libpng 1.6.37 hbc83047_0 libprotobuf 3.11.4 hd408876_0 anaconda libsodium 1.0.17 h516909a_0 conda-forge libstdcxx-ng 9.1.0 hdf63c60_0 libtiff 4.1.0 h2733197_0 libuuid 1.0.3 h1bed415_2 libvpx 1.7.0 h439df22_0 libxcb 1.13 h1bed415_1 libxml2 2.9.9 hea5a465_1 mako 1.1.0 py_0 conda-forge markdown 3.1.1 py36_0 anaconda markupsafe 1.1.1 py36h8c4c3a4_1 conda-forge matplotlib 3.1.3 py36_0 matplotlib-base 3.1.3 py36hef1b27d_0 mistune 0.8.4 py36h8c4c3a4_1001 conda-forge mkl 2020.0 166 mkl-service 2.3.0 py36he904b0f_0 mkl_fft 1.0.15 py36ha843d7b_0 mkl_random 1.1.0 py36hd6b4f25_0 nbconvert 5.6.1 py36h9f0ad1d_1 conda-forge nbformat 5.0.6 py_0 conda-forge ncurses 6.2 he6710b0_0 notebook 6.0.3 py36_0 conda-forge numpy 1.18.1 py36h4f9e942_0 numpy-base 1.18.1 py36hde5b4d6_1 olefile 0.46 py_0 conda-forge opencv 3.4.2 py36h6fd60c2_1 openssl 1.0.2u h516909a_0 conda-forge opt_einsum 3.1.0 py_0 anaconda pandas 1.0.3 py36h830a2c2_1 conda-forge pandoc 2.9.2.1 0 conda-forge pandocfilters 1.4.2 py_1 conda-forge parso 0.7.0 pyh9f0ad1d_0 conda-forge pcre 8.43 he6710b0_0 pexpect 4.8.0 py36h9f0ad1d_1 conda-forge pickleshare 0.7.5 py36h9f0ad1d_1001 conda-forge pillow 7.0.0 py36hb39fc2d_0 pip 20.0.2 py36_1 pixman 0.38.0 h7b6447c_0 prometheus_client 0.7.1 py_0 conda-forge prompt-toolkit 3.0.5 py_0 conda-forge protobuf 3.11.4 py36he6710b0_0 anaconda ptyprocess 0.6.0 py_1001 conda-forge py-opencv 3.4.2 py36hb342d67_1 pygments 2.6.1 py_0 conda-forge pygpu 0.7.6 py36hc1659b7_1000 conda-forge pyparsing 2.4.7 pyh9f0ad1d_0 conda-forge pyqt 5.6.0 py36h13b7fb3_1008 conda-forge pyrsistent 0.16.0 py36h8c4c3a4_0 conda-forge python 3.6.5 hc3d631a_2 python-dateutil 2.8.1 py_0 conda-forge python_abi 3.6 1_cp36m conda-forge pytz 2019.3 py_0 conda-forge pyyaml 5.3.1 py36h8c4c3a4_0 conda-forge pyzmq 19.0.0 py36h9947dbf_1 conda-forge qt 5.6.3 h8bf5577_3 readline 7.0 h7b6447c_5 requests 2.12.5 py36_0 conda-forge scipy 1.4.1 py36h2d22cac_3 conda-forge send2trash 1.5.0 py_0 conda-forge setuptools 46.1.3 py36_0 sip 4.18.1 py36hf484d3e_1000 conda-forge six 1.14.0 py36_0 sqlite 3.31.1 h62c20be_1 tensorboard 1.15.0 pyhb230dea_0 anaconda tensorflow 1.15.0 mkl_py36h4920b83_0 anaconda tensorflow-base 1.15.0 mkl_py36he1670d9_0 anaconda tensorflow-estimator 1.15.1 pyh2649769_0 anaconda termcolor 1.1.0 py36_1 anaconda terminado 0.8.3 py36h9f0ad1d_1 conda-forge testpath 0.4.4 py_0 conda-forge theano 1.0.3 py36hfc679d8_1 conda-forge tk 8.6.8 hbc83047_0 toolchain 2.4.0 0 toolchain_c_linux-64 2.4.0 0 toolchain_cxx_linux-64 2.4.0 0 tornado 6.0.4 py36h8c4c3a4_1 conda-forge tqdm 4.45.0 pyh9f0ad1d_0 conda-forge traitlets 4.3.3 py36h9f0ad1d_1 conda-forge wcwidth 0.1.9 pyh9f0ad1d_0 conda-forge webencodings 0.5.1 py36_1 anaconda werkzeug 0.16.1 py_0 anaconda wheel 0.34.2 py36_0 wrapt 1.12.1 py36h7b6447c_1 anaconda xz 5.2.5 h7b6447c_0 yaml 0.2.4 h516909a_0 conda-forge zeromq 4.3.2 he1b5a44_2 conda-forge zipp 3.1.0 py_0 conda-forge zlib 1.2.11 h7b6447c_3 zstd 1.3.7 h0b5b093_0
May you need the 2nd train(train_iou.py).
RafayAK
commented
4 years ago @gmt710 this issue was not related to train_iou.py. Already solved using the author's suggestion.
gmt710
commented
4 years ago boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0))
ValueError: not enough values to unpack (expected 4, got 3)@RafayAK ,I met this problem too.Because I do not train the 2nd step by train_iou.py and only eval the 1st step model(train.py).
eg4000
commented
4 years ago model = models.load_model(model_path, backbone_name='resnet50',convert=True)
malik-anhar
commented
3 years ago Hi @RafayAK thanks for opening this issue, I met the same problem when I run the model with
boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0))
an error occurred
ValueError: not enough values to unpack (expected 4, got 3)
but it's already solved by adding convert=True while loading the model.
Then I'm trying to retrain the model with my own dataset. The training running well, but when I run prediction with my retrained model and enabling the convert by adding convert=True, but the error still happen ValueError: not enough values to unpack (expected 4, got 3)
Is there something wrong with the training?
*the retrained model can be found here
Hi Eran, I tried to test out the pre-trained model that you graciously provided in #9, but I can't seem to figure out why it does not work? Following is the way I'm trying to make it work:
... this is where the inference fails:
I've been looking at previously closed issues in solving this #21, #51, and #30, sadly to no avail. Any help will be much appreciated
PS: Following are my dependencies setup: