https://github.com/edusense/production/tree/master/backfill

auto_backfill.py may require inspection due to the changes of variables and/or syntax of certain strings.
We need to double-check it's functionality, it should be good to roll but a concern of mine.

Now or with new gaze pipeline but we want to rely on this functionality for scale.

Problem:

While --video_schema and --audio_schema name can be anything when running run_backfill.py there is an underlying assumption that "-audio" and "-video" are on the end of the schema name. run_backfill.py does not automatically append them when running the containers.

Impact:

After successfully inserting videoframes or audioframes into a session, the query to retrieve them is not as expected or impossible to form. If the original schema appended "-audio" or "-video" (eg. schema: "edusense-audio") then only using the first part of the schema will succeed. If the original schema did not append this then the frames are not queryable (eg. schema: "0.1.0" transforms to "0.1.0-audio").

Solution:

In storage/query/resolver/session.go lines 123 and 165 need to have the second argument updated to remove the string concatenation.

Side effects:

Any arguments automatically passed into scripts like run_backfill.py will need to be updated to match the fully qualified expected schema.

The getting started guide mentions

Frontend (not included in this repository)

• frontend visualizes processed data, providing useful debug interface. Currently, our edusense repository does not contain this component.

are there plans to open source the front end as well?

Python2 will reach its end of life by the end of 2019. Updated video pipeline code and Dockerfile accordingly to support python3.

https://github.com/edusense/edusense/tree/master/compose

@pranavdheer @prasoonpatidar @nvogety
Sending to Jameel as he was running into issues.

The functionality seems to be there but no docker-compose.audio.yml.

docker run edusense/audio --help
usage: audio_pipeline.py [-h] --front_url [FRONT_URL] --back_url [BACK_URL]
[--backend_url [BACKEND_URL]]
[--time_duration [TIME_DURATION]]
[--session_id [SESSION_ID]] [--schema [SCHEMA]]

EduSense audio pipeline

optional arguments:
-h, --help show this help message and exit
--front_url [FRONT_URL]
URL to rtsp front camera stream
--back_url [BACK_URL]
URL to rtsp back camera stream
--backend_url [BACKEND_URL]
EduSense backend
--time_duration [TIME_DURATION]
Set the time duration for file to run
--session_id [SESSION_ID]
EduSense session ID
--schema [SCHEMA] EduSense schema

and the following branch documentation for docker-compose needs to be updated from:

docker run edusensecmu/video --help
docker run edusensecmu/audio --help
docker run edusensecmu/openpose --help
docker run edusensecmu/storage --help

To:
docker run edusense/video --help
docker run edusense/audio --help
docker run edusense/openpose --help
docker run edusense/storage --help

• Add images
• Put more details, etc

We have openpose v1.5.1 submodule on our codebase, but recent release of Openpose v1.6.0 breaks our Docker build.

Error log:

Step 16/28 : RUN cmake /openpose && make -j 8 && make install
**OMITTED**
-- Found glog    (include: /usr/include, library: /usr/lib/x86_64-linux-gnu/libglog.so)
-- Found Protobuf: /usr/lib/x86_64-linux-gnu/libprotobuf.so;-lpthread (found version "3.0.0")
-- Found OpenCV: /usr (found version "3.2.0")
-- Caffe will be downloaded from source now. NOTE: This process might take several minutes depending
        on your internet connection.
-- Caffe has already been downloaded.
fatal: not a git repository: /openpose/3rdparty/caffe/../../../../../../.git/modules/compute/openpose/third_party/openpose/modules/3rdparty/caffe
-- Caffe will be built from source now.
-- Download the models.
-- Downloading BODY_25 model...
-- NOTE: This process might take several minutes depending on your internet connection.
CMake Error at cmake/Utils.cmake:8 (file):
  file DOWNLOAD HASH mismatch
    for file: [/openpose/models/pose/body_25/pose_iter_584000.caffemodel]
      expected hash: [78287b57cf85fa89c03f1393d368e5b7]
        actual hash: [d41d8cd98f00b204e9800998ecf8427e]
             status: [22;"HTTP response code said error"]
Call Stack (most recent call first):
  CMakeLists.txt:976 (download_model)
-- Not downloading body (COCO) model
-- Not downloading body (MPI) model
-- Downloading face model...
-- NOTE: This process might take several minutes depending on your internet connection.
CMake Error at cmake/Utils.cmake:8 (file):
  file DOWNLOAD HASH mismatch
    for file: [/openpose/models/face/pose_iter_116000.caffemodel]
      expected hash: [e747180d728fa4e4418c465828384333]
        actual hash: [d41d8cd98f00b204e9800998ecf8427e]
             status: [22;"HTTP response code said error"]
Call Stack (most recent call first):
  CMakeLists.txt:982 (download_model)
-- Downloading hand model...
-- NOTE: This process might take several minutes depending on your internet connection.
-- Models Downloaded.
CMake Error at cmake/Utils.cmake:8 (file):
  file DOWNLOAD HASH mismatch
    for file: [/openpose/models/hand/pose_iter_102000.caffemodel]
      expected hash: [a82cfc3fea7c62f159e11bd3674c1531]
        actual hash: [d41d8cd98f00b204e9800998ecf8427e]
             status: [22;"HTTP response code said error"]
Call Stack (most recent call first):
  CMakeLists.txt:984 (download_model)
-- Configuring incomplete, errors occurred!
See also "/app/build/CMakeFiles/CMakeOutput.log".
See also "/app/build/CMakeFiles/CMakeError.log".
The command '/bin/sh -c cmake /openpose && make -j 8 && make install' returned a non-zero code: 1

Potential Solution

There are two potential solutions:

1. update submodule to openpose follow 1.6.0 and update our patch files at https://github.com/edusense/edusense/tree/master/compute/openpose/edusense. You may find some guidance at https://github.com/edusense/edusense/blob/master/compute/openpose/README.md to figure out how to create a patch file, etc.
2. figure out whether openpose team allows us to download parameter files at specific version. (namely, v1.5.0)