OBJECTIVE
-------------------
Measure the people flow, face detection and attractivity of physical showcases (offline stores) through cameras. The result will be a selling funnel an overview with the best days and times of the business and the customer profile with gender classification and age estimative.
DEMMAND
------------------
Develop a multi platform algorithm to solve the mentioned objective with the below structure or another equivalent
NAME: getDoDAnalysis
INPUT 1: videoToAnalyse // Real time streaming url, youtube video or a local video file
output 1: dodevents // database table with captured events. Table structure below.
OUTPUT 2: analisedVideo // Marked video with events (face bounding box, gender, and age) in an MP4 file or real time streaming
OBSERVATIONS
--------------------------
1. We don't have technology restriction, but we are looking for a method or technique with the best precision
2. The gender and age identification just is needed in the cases that the person stopped in front of the showcase, not for all passers
3.
The event recording of passengers must happen, in this case without gender and age fields
4. The face recognize must happen when the passenger stop for at least 4 seconds in front of the showcase or enter into the store
5. All the events must be recorded with precision of milliseconds
6.
The delivery must include the algorithms related the training of face detection, gender classification, age estimative and people detection
7. We can offer access to a remote server with adequate capacity to support the neural network training
8. Our framework suggestion is Tensorflow
business results
--------------------------------
with the collected events, we will answer questions like:
a) how many people passed in front the store today?
b) how many men stopped to see the showcase this month?
c) how many people entered the store (crossed the passage line)?
d) what's the average time that the people stay in front of the showcase?
e) how many people up to 30 years old viewed the showcase this week?
f) how many people with over than 50 years entered the store this year?
scenarios
-------------------
take a look at the events table simulation.
The project will be applyed in store scenarios with many cameras, in these examples on the events table we used just 3, in front of the store, inside and with the focus on check out.
1. Analysis of one person with selling funnel using 3 cameras
2. One person passed in front of the store and didn't stop to see the showcase (without face detection).
3. Two people entered simultaneously into store, talked with the seller and finished the buy at the checkout
public cameras
------------------------------
security cameras with examples that we will work with (user admin without password):
http://77.243.103.105:8081/view/
viewer_index.shtml?id=3274
http://177.81.27.91:60001/
view2.html
http://189.120.175.77:60001/
view2.html
http://131.95.3.162/view/
viewer_index.shtml?id=848
http://177.35.158.108:60001/
view2.html?type=main
http://82.127.3.227:8081/view/
viewer_index.shtml?id=307
From this project, we can start a nice and long term partnership with the developer that work with us. Welcome aboard! :)
Delivery term: Not specified