[bug] fix bugs in running demo inference

[LXXXXR]

No description provided.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

lixinran seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov]

Codecov Report

Merging #71 into master will decrease coverage by 0.04%.
The diff coverage is 33.33%.

@@            Coverage Diff             @@
##           master      #71      +/-   ##
==========================================
- Coverage   67.68%   67.63%   -0.05%     
==========================================
  Files          71       71              
  Lines        3085     3087       +2     
  Branches      562      562              
==========================================
  Hits         2088     2088              
- Misses        949      951       +2     
  Partials       48       48              

Flag	Coverage Δ
#unittests	67.63% <33.33%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmcls/apis/inference.py	22.22% <33.33%> (-0.86%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 784987f...b2051a9. Read the comment docs.

[ycxioooong]

We should support the input parameter img to be either np.ndarry or str.
So to improve the code:

1. in docstring: change the description of img
2. use isinstance(...) to check the type of img. If it is str, load image and execute pipeline; else, use another pipeline.

[yl-1993]

It seems that test_pipeline of configs, e.g. imagenet_bs32, has already contained LoadImageFromFile?