VIDEO TESTING REPORT
1. Executive Summary
This report summarizes the findings from testing the video emotion analysis feature. Several test
scenarios were conducted to evaluate the system's accuracy, functionality, and user experience under
varying conditions.
2. Test Cases and Observations
- In general, the campaign was created and sent successfully to different emails at the same time.
However, it takes time for the link to be sent to the emails.
Test Case 1: Video with No Audio (10 Seconds)
• Scenario: Video with no audio tested 10 times.
• Results:
o Accurate results: 7 tests user is "happy," 3 tests user is "sad." The model predicts
the highest chance of the results is happy.
o Result was sent successfully to the database, and successfully created the report.
o Issues:
- In some cases, the model predicts surprise, anger, fear, which is
inaccurate.
- The voice of user is echoing when watching video.
• Notes: System incorrectly classifies significant emotions occasionally. For emotions such
as angry, fear, surprise, the model predicts inaccurately most of the time.
Test Case 2: Video with Audio (1 Minute)
• Scenario: Video with audio tested for 1 minute.
• Results:
o General results are good.
o Significant emotion predictions (surprise, fear, angry, disgust) remain inaccurate.
o Echoing audio impacts usability.
Test Case 3: Medium Lighting Conditions
• Scenario: Video tested under medium lighting.
• Results:
o Works in most cases.
o Fails in darker conditions.
• Notes: Lack of lighting warnings or instructions for users to ensure adequate lighting.
Test Case 4: Dropped Video (Incomplete)
• Scenario: Video drops halfway through.
• Results:
o No result is sent.
• Notes: Recommend adding a completion rate chart to indicate video progress and
improve user feedback.
•
Test Case 5: Incomplete Test
• Scenario: Video finished but no recording or form submission.
• Results:
o Charts still display tone and emotions analysis results, leading to inaccurate or
misleading data.
Test Case 6: User Experience Issues with Recording
• Scenario: 4 users do the test from their computer.
• Results:
o Major concern: After watching the test, other 4 users cannot process more to
recording steps. The website does not jump to recording steps.
o Some users are hesitant to grant camera access due to privacy concerns.
o Automatic recording starts after the video ends, but users often do not realize this,
leading to incomplete recordings.
o Some users attempt to skip ads.
o The web test was designed not really attractive, so users got distracted.
• Notes:
- Recommend adding a prominent "Next Step" button and clearer
instructions for recording. Highlight when recording starts and provide
optional prompts to reassure users about privacy.
- Have some methods to block users skipping ads.
- Improve the visualization
3. Gaps in testing
While the current testing covered various scenarios, the following areas remain untested:
1. Mobile Testing:
• The platform has not been tested on mobile devices to ensure responsiveness and
functionality.
2. Scalability with Larger User Groups:
• Testing with a significant number of participants (e.g., hundreds or thousands) has not
been conducted.
3. Comparison of Personal and General Reports:
• Testing has not explored the generation and comparison of personalized vs general
reports.
4. Edge Cases for Video Input:
• Videos with extreme conditions, such as very fast facial movements, poor resolution, or
heavy background noise.
• Videos with multiple faces in the frame. Does the system focus on one face, or does it
average the emotions?
5. Stress Testing:
• Evaluate system performance under extreme conditions, such as simultaneous uploads by
multiple users or very large video files.
6. Error Handling and Recovery:
• Simulate errors, such as network interruptions during uploads or data processing, to
check how gracefully the system recovers.
