Release Notes
Current deep learning model: v2.18.8(10 June 2024)
June 10, 2024
Python cochl
library has been released for easier API integration.
The new model v2.18.8
has been deployed.
December 28, 2023
Cochl.Sense Cloud API v1.4.0
released
- Implemented a “Default Hop size” feature enabling users to set the inference interval at either 0.5 seconds or 1 second.
- Introduced a new “Sensitivity Control” feature that allows users to adjust the recognition sensitivity for each tag.
- Updated sample codes now include a “Result Summary” feature for a simplified display.
June 21, 2023
Model version updated to v2.15.8
- Model performances improved.
- Number of detectable sound tags increased to 103 sounds.
Others
tag probability changed to constant ‘0.0’.
September 28, 2022
Model version updated to v2.7.9
The ‘siren’ tag has been modified and split into 2 tags :
Civil_defense_siren
Emergency_vehicle_siren
Other tags have been added.
July 07, 2022
Two new concepts have been implemented :
multi-tag
: multiple sounds can be recognized at the same time.no service
: all the sounds can be inferenced regardless of their category : no need to specify a service anymore.
Breaking changes
Due to multi-tag concept, the ‘Read status’ request response format has been modified.
March 15, 2022
Two new parameters are available from client side when creating a session :
service
: set service to inference different events with same project / api keymetadata
: those metadata will later be available on analytics dashboard.
Breaking changes
Events for each service have been refactored.
Aug. 20, 2021
Breaking changes
For better consistency, some event names have been renamed.
- ‘Glassbreak’ becomes ‘Glass_break’
- ‘Clap becomes’ ‘Double_clap’
- ‘Whistling’ becomes ‘Whistle’
- ‘Burping’ becomes ‘Burp’
- ‘Snoring’ becomes ‘Snore’
Some event recognition have been removed.
- Whisper
- Keyboard_or_mouse
Apr. 21, 2021
- Smart filtering is now always activated. Changing smart-filtering option on client will not have any effect.
Nov. 27, 2020
Pricing
Decreasing API price from $0.048 /min to $0.012 /min
Company
- Our new domain has change from cochlear.ai to cochl.ai
- A brand new site accessible at https://www.cochl.ai
Model
Addition of Human Status service:
- Baby_cry
- Dining_clink
- Dog_bark
- Electric_shaver_or_toothbrush
- Keyboard_or_mouse
- Knock
- Toilet_flush
- Water_tap_or_liquid_fill
- Others
Sept. 07, 2020
Client
Add smart filtering capability to all clients v1.1.0.
Breaking changes
Java and Dart dependency name have change.
For java, now use
dependencies {
implementation 'ai.cochlear.sense:cochl-sense:1.1.0'
}
For dart, now use
dependencies:
cochl_sense: ^1.1.0
Jul. 21, 2020
Model
Addition of Human status service:
- Burping
- Cough
- Fart
- Hiccup
- Laughter
- Sigh
- Sneeze
- Snoring
- Yawn
- Others
May 15, 2020
Official v1 release
Model
- Better accuracy
- Faster result
- Specialized model to 2 different services :
- Emergency Detection: Fire/smoke alarm, glassbreak, scream, and siren
- Human Interaction: Clap, finger snap, knock, whisper, and whistling
Client
- Dart client
- Clients use SSL to send encrypted audio data
Breaking changes
- Change client API
- Split supported sounds between different services
- Stop support of beta API key
- All services and API keys from beta version are not supported anymore
- All beta clients needs to be updated to latest version
Dashboard
- Creation of dashboard: https://dashboard.cochl.ai
- Data usage visualization
- Usage of multiple API Key using different projects for better control
Jan. 22, 2020
Release of beta v2.1 : number of supported sounds increased to 34 classes.
Model
Added (14 classes)
- Bicycle_bell
- Birds
- Burping
- Cat_meow
- Clap
- Crowd_applause
- Crowd_scream
- Explosion
- Finger_snap
- Keyboard_mouse
- Mosquito
- Sigh
- Whisper
- Wind_noise
Changed (3 classes)
- Applause → Crowd_applause
- Gunshot_explosion → Gunshot
- Scream_shout → Scream
Deleted (1 class) :
- Baby_laughter
Jul. 03, 2019
Client
- Python: Fix time frame bug in streaming function.
Jun. 05, 2019
Cochl.Sense beta v2 is now available.
Model
Acoustic event API
- Removed the subtask option.
- Available sound events are increased to 22 (v-1.0: 7 events)
- Provides API with a new serving system which is more scalable.
Music API, Speech API
- Paused temporarily and will be back in Q3 2019 with improved performance.
Client
- Protobuf messages changed
- gRPC service changed
Nov. 09, 2018
Cochl.Sense beta v1 is now available in Beta release, which is the first general purpose audio cognition API. The API is designed to empower developers and corporates with Machine Listening AI. During Beta period, it is free to use within the daily quota of 700 calls (for file inputs) and 10 minutes (for streaming).
Model
- Speech and music activity detection
- Age and gender detection
- Additional sound event class: glassbreak
- Improved performance
Client
- Improved latency
- Streaming input support
- Example client codes of other languages (Java, Node.js)
- Additional functionalities