I have the exact same problem but haven’t found a good solution, if you use roboflow==0.2.23 it will instead download the json_response file with another error. However I found that by doing the image recognition online it seams to work if that is viable to you? Check out my post to see if you find anything of help and please let me know if you solve it! Roboflow Post
Here is a way to run It online:
Install inference.
pip install inference
If you have an NVIDIA GPU that supports CUDA
pip install inference-gpu
Add your Roboflow API Key as an environment variable or in a .env file. Just run it in the terminal basically
export ROBOFLOW_API_KEY="YOUR_PRIVATE_API_KEY"
Use it in your python project.
import inference
model = inference.get_model("Project_ID/Version number") #It should be in one sentence like this (untilted-asfg/1) where untilted-asfg is the model ID and 1 is the version
results = model.infer(image="YOUR_IMAGE.jpg")
print(results)
please let me know if that works for you! If my "tutorial" is uncomprehendable you can just find it under model-> deployments -> Self Hosted Reference (native Python)