I found and fixed some Wrong in README when I follow it on a clear Unbuntu OS

[libingwei]

I found three point as follows :

1. The pandas version should be 0.24.0 int step 1.

2. The command "pip3 install $(python3 util/taskcluster.py --decoder)" should be placed after "git clone https://github.com/mozilla/DeepSpeech.git" .

3. The "attack.py " should be replaced by " classify.py " in step 5 and step 8.

I modified it as follows：

(Notice the bold parts)

---

Instructions for basic use:

1. Install the dependencies pip3 install tensorflow-gpu==1.14 progressbar numpy scipy pandas==0.24.0 python_speech_features tables attrdict pyxdg

• 1a. Make sure you have installed git lfs. Otherwise later steps will mysteriously fail. Download and install : https://git-lfs.github.com/

2. Clone the Mozilla DeepSpeech repository into a folder called DeepSpeech: git clone https://github.com/mozilla/DeepSpeech.git

• 2a. Install decoder cd DeepSpeech pip3 install $(python3 util/taskcluster.py --decoder)

• 2b. Checkout the correct version of the code: git checkout tags/v0.4.1

• 2c. If you get an error with tflite_convert, comment out DeepSpeech.py Line 21 # from tensorflow.contrib.lite.python import tflite_convert

3. Download the DeepSpeech model wget https://github.com/mozilla/DeepSpeech/releases/download/v0.4.1/deepspeech-0.4.1-checkpoint.tar.gz tar -xzf deepspeech-0.4.1-checkpoint.tar.gz

4. Verify that you have a file deepspeech-0.4.1-checkpoint/model.v0.4.1.data-00000-of-00001, Its MD5 sum should be ca825ad95066b10f5e080db8cb24b165 md5sum deepspeech-0.4.1-checkpoint/model.v0.4.1.data-00000-of-00001

5. Check that you can classify normal audio files correctly python3 classify.py --in sample-000000.wav --restore_path deepspeech-0.4.1-checkpoint/model.v0.4.1

6. Generate adversarial examples python3 attack.py --in sample-000000.wav --target "this is a test" --out adv.wav --iterations 1000 --restore_path deepspeech-0.4.1-checkpoint/model.v0.4.1

7. Verify the attack succeeded python3 classify.py --in adv.wav --restore_path deepspeech-0.4.1-checkpoint/model.v0.4.1

---

[tom-doerr]

Additionally, "images" in step 5 could be changed to "audio files".

[carlini]

Ah yes, thanks for catching this. I will make these changes.