Hello All;

Hope your are doing well;

In Mask RCNN I would like change the color of the mask to be White with the Alpha = 1, I change it frome the right place in vizualize.py, but anything change the mask color still red or blue or another color, why the changes on vizualize.py dont have effect ?

Is there an explanation for what these parameters represent?

octave_base_scale
scales_per_octave
anchor_ratios
anchor_strides
featmap_strides

And how can I calculate the best ones for my data? (Which contains lots of very small objects)

I understand that these two python files show two different methods to construct a model. The original n_epoch is 500 which works perfect for both python files. But if I change n_epoch to 20, only tutorial_mnist_mlp_static.py can achieve a high test accuracy (~0.97). The other file tutorial_mnist_mlp_static_2.py only get 0.47.

The models built from these two files looks the same for me (the s

I created a virtual environment using conda and I installed darkflow and it worked pretty well. I later exported the environment to try it on another computer to verify that everything works. But conda doesn't install the darkflow even though it exists on the requirement.txt

A more consistent and multi-functional global level of verbosity control,
suggest an enhancement that will see print(...) in project be converted to using the python logging. module

import logging
#Then instead of print() use either
logging.info(......)
#or
logging.debug(.....)
#or
logging.warning(....)
#or
#logging.error()

In that way verbosity can be globally

To get started using this repo quickly using a Google Cloud Platform (GCP) Deep Learning Virtual Machine (VM) follow the instructions below. New GCP users are eligible for a $300 free credit offer. Other quickstart options for this repo include our [Google Colab Notebook](https://colab.research.google.com/github/ultralytics/yolov3/blob

Hi, thanks for the great code!

I wonder do you have plans to support resuming from checkpoints for classification? As we all know, in terms of training ImageNet, the training process is really long and it can be interrupted somehow, but I haven't notice any code related to "resume" in scripts/classification/train_imagenet.py.

Maybe @hetong007 ? Thanks in advance.

Per my understanding, the functions bunched together in the sub-directory tf_extended are meant to supplement the SSD implementation using standard TensorFlow functions, but it is not the same as TFX - Tensorflow Extended. Is this correct?

If that's the case, perhaps a modification to the readme will help newcomers avoid conflating the two. I'm willing to

Damn. Is there any one who can clearly explain below two lines???
link

matches = truths[best_truth_idx]          # Shape: [num_priors,4]
conf = labels[best_truth_idx] + 1         # Shape: [num_priors]

When I delete the lib.utils. ,the test.py is right but the demo.py is wrong , however,when I use the lib.utils. , the demo.py is right but the test.py is wrong !!

why? why

When attempting to download cityscapes_2048x1024 I got: ./download-models.sh: line 721: download_fcn_resnet18_cityscapes_2048x512: command not found
It looks like there was a typo, and line 721 needs to be changed from:
download_fcn_resnet18_cityscapes_2048x512 to download_fcn_resnet18_cityscapes_2048x1024

Thanks for the amazing repo!

Can you describe what modifications need to be done if I want to replace dynamic_rnn with tf.keras.RNN in many-to-one example as dynamic_rnn is deprecated now.

hello大家好，我是该仓库的作者。鉴于我复现tensorflow-yolov3踩了太多坑，特此发个贴，帮助大家少走弯路。大家有问题可以在下面留言。

does simpledet support for negative examples without any foreground instances? and there is no
relative description in the dataset.md.

• Explain in notebook/FAQ what non-maxima suppression is what values to set (threshold on IoU)
• Explain and provide code how to pick a good score threshold (reuse Patrick's plot which was implemented for the drone demo)

Describe the bug
Project settings -> tags, are not saved unless user hits 'Enter' after typing new tag.

To Reproduce
Steps to reproduce the behavior:

1. In project settings, add new tag.
2. Click 'Save Project'.
3. Exit project settings pane and return.
4. New tag is not saved.

Expected behavior
Tag should be saved when clicking 'Save Project'.

Screenshots
If a

https://github.com/google/automl/blob/3614751749a21ca2fcb299b60238c6651ff51125/efficientdet/efficientdet_arch.py#L402

When survival_prob=0.0, in both class_net and box_net, neither dropout or residual connection is applied. While in MBConvBlock, dropout is not applied but residual connection is still present. Did I get it right?